Astrocytic GLUT1 deletion in adult mice enhances glucose metabolism and resilience to stroke

Abstract

Brain activity relies on a steady supply of blood glucose. Astrocytes express glucose transporter 1 (GLUT1), considered their primary route for glucose uptake to sustain metabolic and antioxidant support for neurons. While GLUT1 deficiency causes severe developmental impairments, its role in adult astrocytes remains unclear. Here, we show that astrocytes and neurons tolerate the inducible, astrocyte-specific deletion of GLUT1 in adulthood. Sensorimotor and memory functions remain intact in male GLUT1 cKO mice, indicating that GLUT1 loss does not impair behavior. Despite GLUT1 loss, two-photon glucose sensor imaging reveals that astrocytes maintain normal resting glucose levels but exhibit a more than two-fold increase in glucose consumption, indicating enhanced metabolic activity. Notably, male GLUT1 cKO mice display reduced infarct volumes following stroke, suggesting a neuroprotective effect of increased astrocytic glucose metabolism. Our findings reveal metabolic adaptability in astrocytes, ensuring glucose uptake and neuronal support despite the absence of their primary transporter.

Fig. 1. Inducible deletion of astrocytic GLUT1 in adult mice.

a Generation of GLUT1^fl/fl;GLAST^CreERT2/+ (GLUT1 cKO) mice and littermate controls (GLUT1^fl/fl;GLAST^+/+). b Tamoxifen injections over 5 consecutive days in 8- to 10-week-old mice, with analyses conducted 60 days post-injection. c, d Western blot of GLUT1 (45 kDa) in capillary-depleted brain tissue shows a 48 ± 5% reduction in cKO mice (n = 4) compared to controls (n = 5, p = 0.0005, two-sided unpaired t-test). Vinculin served as loading control. Immunolabeling for GLUT1 and glutamine synthetase (GS) in cortex (e) and hippocampus (f) shows a marked reduction of GLUT1 signal in cKO brain sections (bottom panels). White arrows highlight GLUT1-depleted astrocytes, while capillaries (arrowheads) retain GLUT1 expression. Observed in four mice per genotype. g Cre-dependent AAV (PHP.eB-hGFAP-DIO-EGFPL10a) injected intravenously into GLUT1^fl/fl;GLAST^CreERT2/+ mice (cKO) and GLUT1^+/+;GLAST^CreERT2/+ mice (as controls) enabled astrocyte-specific EGFP-L10 expression for translating ribosome affinity purification (TRAP). Tamoxifen was administered 7 days post-AAV injection, and TRAP RNA was collected 60 days later. h Immunostaining confirms widespread EGFP-L10 expression (anti-GFP) in astrocytes (anti-S100β) using the approach in (g), observed in two mice per genotype. i qPCR analysis of astrocytic Gfap polysome-associated RNA following TRAP shows an 8.5-fold enrichment in immunoprecipitated (IP) samples vs. input in control (n = 5) and cKO mice (n = 5, p = 0.9873, two-sided unpaired t-test). j qPCR analysis of GLUT1 (Slc2a1) polysome-associated RNA in IP samples shows a 69 ± 5% reduction in cKO mice (n = 5) vs. controls (n = 5, p < 0.0001, two-sided unpaired t-test). k RNA-seq reads (counts per million, CPM) from exons 3–8 of Slc2a1 show a 68 ± 4% reduction in cKO cortical IP samples (n = 4) vs. controls (n = 3, p < 0.0001, two-sided unpaired t test). l RNA-seq analysis of glucose and monocarboxylate transporters reveals no significant differences between genotypes (n = 3 vs. 4, p > 0.85 for all genes, two-way Anova with Šídák’s multiple comparisons test). Data are presented as scatter dot plots with mean ± SEM. Source data are provided as a Source Data file.

Fig. 2. Astrocytic GLUT1 deletion does not affect sensorimotor, learning or memory functions.

a Experimental timeline: Behavioral tests were conducted 60, 75, and 90 days post-tamoxifen treatment, with a one-week rest between tests. b Sensorimotor performance (gait, ledge, hindlimb clasping, and horizontal wire test) was comparable between Ctrl (n = 11) and cKO (n = 11) mice (p = 0.5295, two-sided unpaired t-test). c Left: Barnes maze setup with 22 holes, one escape hole, and visual cues. Right: Experimental timeline with a 4-day acquisition phase (arrowheads indicate sessions), followed by a day 5 memory test with the target hole sealed. d Both Ctrl (n = 11) and cKO (n = 11) mice showed a significant decrease in latency to find the target hole from day 1 to day 4 (Ctrl: β = -93.71 s, cKO: β = −91.05 s, each p < 0.0001, two-sided paired t-test), with no genotype differences in spatial learning (F_interaction(3,60) = 0.6990, p = 0.5563, two-way ANOVA). e On day 5, Ctrl (n = 11) and cKO (n = 11) mice preferentially identified the target hole over an adjacent hole (Ctrl: β = 3.6, p = 0.0027; cKO: β = 4.3, p = 0.0003, two-sided paired t-test), indicating intact memory retrieval in cKO mice. f Locomotor performance on day 5 was comparable between Ctrl (7.9 ± 0.5 m; n = 11) and cKO (7.5 ± 0.5 m; n = 11, p = 0.5979, two-sided unpaired t-test). g Passive avoidance setup with bright and dark compartments. Memory retention was measured by latency to enter the dark compartment 1 h and 24 h post-acquisition. Both Ctrl (n = 11) and cKO (n = 10) mice showed increased latency at 1 h (Ctrl: β = 52.27 s, p = 0.0015; cKO: β = 54.90 s, p = 0.0126) and 24 h (Ctrl: β = 27.09 s, p = 0.058; cKO: β = 22.10 s, p = 0.0040), with no genotype differences (F_interaction(2,38) = 0.08586, p = 0.9179, two-way ANOVA). Data are represented as mean ± SEM. Source data are provided as a Source Data file.

Fig. 3. Astrocytes have normal basal glucose levels in GLUT1 cKO mice.

a Experimental timeline: Intracortical injection of AAVs encoding glucose sensor FLII12 Pglu700μΔ6 (FLIIP) and DIO-TdTomato into ~8-week-old GLUT1^fl/fl;GLAST^CreERT2/+ (cKO) and GLUT1^+/+;GLAST^CreERT2/+ (Ctrl) mice, followed by tamoxifen treatment 7 days later. Acute cortical slice two-photon (2 P) imaging was performed ~60 days post-tamoxifen treatment. b Example 2P images of cortical astrocytes co-expressing FLIIP and TdTomato (marking cells targeted for Cre recombination) selected for glucose imaging (observed in 5 Ctrl and 6 cKO mice). Scale bar: 10 µm. c Representative color-coded glucose sensor ratio images from Ctrl and cKO astrocytes in ACSF with 5 mM glucose ([Glc]_O) and after removal of extracellular glucose (0 mM [Glc]_O). Warm and cold colors indicate high and low glucose levels, respectively. d Quantification of raw sensor ratios from astrocytes under conditions in (c). At 5 mM [Glc]_O and 0 mM [Glc]_O, glucose sensor ratios were comparable between Ctrl (n = 63 cells from 9 slices, 5 mice) and cKO (n = 101 cells from 12 slices, 6 mice) (p = 0.536 at 5 mM [Glc]_O; p = 0.690 at 0 mM [Glc]_O, two-sided linear mixed model). e Normalized glucose sensor ratios (5 mM [Glc]_O ratios normalized to the average minimum at 0 mM [Glc]_O) show comparable basal glucose levels between genotypes (p = 0.831, two-sided linear mixed model). Box plots show the median (center line), quartiles (box bounds), mean (+) and min-to-max (whiskers). Source data are provided as a Source Data file.

Fig. 4. GLUT1 cKO astrocytes have enhanced glucose metabolism.

a Representative glucose sensor traces showing the astrocytic glucose level increase in response to transient elevation of [Glc]_O from 5 mM to 25 mM in Ctrl and cKO brain slices. b Glucose rise rate (dashed lines in (a)) was similar between Ctrl (n = 53 cells from 9 slices, 5 animals) and cKO (n = 88 cells from 12 slices, 6 animals, p = 0.967, two-sided linear mixed model). c Glucose increase (Δ[Glc]) was significantly higher in cKO astrocytes (n = 88 cells from 12 slices, 6 animals) compared to Ctrl (n = 53 cells from 9 slices, 5 animals, p < 0.0001, two-sided linear mixed model). d Schematic of glucose consumption assay: Glucose uptake was inhibited with the GLUT blocker cytochalasin B (CytoB) to assess glucose consumption rate in astrocytes. Glc-6P, glucose-6-phosphate. e Representative traces showing the astrocytic glucose decline in Ctrl and cKO brain slices during CytoB incubation. f Glucose consumption rate, measured as the decline in [Glc] (dashed lines in (e)), was 2.6-fold higher in cKO astrocytes (−0.079 ± 0.003 min⁻¹, n = 101 cells from 12 slices, 6 animals) compared to Ctrl (n = 63 cells from 9 slices, 5 animals, −0.030 ± 0.003 min⁻¹, p = 0.002, two-sided linear mixed model). Box plots show the median (center line), quartiles (box bounds), mean (+) and min-to-max (whiskers). Source data are provided as a Source Data file.

Fig. 5. Reduced stroke-induced neural injury in GLUT1 cKO mice.

a Experimental timeline: Stroke was induced in male mice 60 days post-tamoxifen, followed by perfusion 7 days later. MCA, middle cerebral artery; LSI, laser speckle imaging. b Representative (LSI) images of Ctrl and cKO mice pre- and post-stroke, with infarct areas outlined. c Cerebral blood flow (CBF) changes (%) (left) and hypoperfused area size (right) showed no significant differences between Ctrl (n = 9) and cKO (n = 8, p = 0.3353 and p = 0.8724, two-sided unpaired t-tests). d Post-stroke body weight changes did not differ between genotypes (n = 9 vs. 8, F_interaction(4,60) = 2.436, p = 0.7515, two-way ANOVA). e NeuN (neurons), GFAP (astrocytes) and DAPI (nuclei) stainings of infarcted brain sections 7 days post-stroke. Dashed outlines indicate lesion areas. f Neural injury volumes (in mm³) were significantly reduced in cKO (n = 8) vs. Ctrl (n = 9) when quantified using NeuN and GFAP (p = 0.011, linear mixed model). Individually, NeuN-based injury volume was 9.91 mm³ in Ctrl and 5.50 mm³ in cKO, while GFAP-based volumes were 9.02 mm³ in Ctrl and 5.26 mm³ in cKO. On average, cKO mice had 43.1% ± 8.4% SEM smaller lesions. Box plots show the median (center line), quartiles (box bounds), mean (+), and 1.5× interquartile range (whiskers). g Immunolabelling of NeuN, GFAP, IBA1 (microglia), and DAPI at the infarct. Dotted lines indicate infarct border and peri-infarct region (observed in 7-8 mice per genotype). h Sholl analysis of infarct border astrocytes showed a comparable number of intersections between genotypes, with a significant difference observed at 76 µm from the soma (n = 48 vs. 56 cells from 7 and 8 animals, p = 0.03; linear mixed-effects model with post-hoc pairwise comparisons). i Sholl analysis of peri-infarct astrocytes revealed a significantly fewer intersections in cKO compared to Ctrl (n = 97 vs. 110 cells from 7 and 8 animals, p < 0.001 within 16–30 µm from the soma; linear mixed model with post-hoc pairwise comparisons). Data in (c, d, h, and i) are represented as mean ± SEM. Source data are provided as a Source Data file.