Improvement in Insulin Sensitivity Prevents Decline in Glucose Uptake, Functional Connectivity, and Volume in the Insulin Resistant Human Brain

Abstract

Insulin resistance (IR) is a modifiable risk factor for dementia, yet its effects on brain metabolism and function remain unclear. In older adults, greater IR was associated with reduced cerebral glucose uptake (indicating impaired mitochondrial metabolism), atrophy, and weakened connectivity between brain regions critical for cognition. In neuron-specific insulin receptor knockout mice, brain IR produced deficits in hippocampal- and prefrontal-dependent tasks accompanied by reduced brain mitochondrial ATP and elevated reactive oxygen species. To evaluate reversibility of IR-induced brain deficits, forty older adults with IR were randomized to 40-weeks of metformin or placebo. Metformin improved insulin sensitivity, increased brain glucose uptake, strengthened cognitive network connectivity, and preserved whole-brain and regional volumes implicated in decision-making and learning. Metformin also improved processing speed and working memory. Collectively, these findings highlight IR as a driver of brain metabolism and support the concept that insulin sensitization can prevent neurobiological deficits in older people with IR.

Figure 1. People with Whole Body Insulin Resistance Display Lower Glucose Uptake in Brain Regions Critical for Executive Function, Sensory Integration, and Attention.

(A) Standardized 3-dimensional FDG SSP (Stereotactic Surface Projection) intensity maps demonstrate whole brain glucose uptake across ranging levels of fasting glucose. (B-G) X-Y plot demonstrate the significant negative correlation between fasting glucose and glucose uptake in the whole brain, middle frontal gyrus, pars opercularis, inferior parietal lobule, angular gyrus, and posterior cingulate cortex. Sample size contributing to linear regression analysis between brain glucose uptake and fasting plasma glucose was n=70 (Healthy Control, HC=17, green dots; Insulin Resistant, IR=53, purple dots). (H) Group differences in HbA1c, controlled fasting glucose, insulin, C-peptide, and HOMA-IR. Sample sizes were n=18 for HC and n=56 for IR for all outcomes except for controlled fasting C-peptide (IR=55) and HbA1c (IR=54). (I-J) Glucose uptake in the whole brain (I) and in different brain regions (J) measured by ¹⁸FDG-PET and displayed as relative standardized uptake values (SUVr) (HC, n=17; IR, n=55). Dots represent individual participants. Simple linear regression analysis was used for panels in B-G. Two-sided Student’s t-tests were used to compare group differences for panels in H-J.

Figure 2. Brain Regions Important for Executive Function and Memory have Fewer Positive Functional Connections in People with Insulin Resistance.

(A) Chord diagram represents all significant (FDR<0.05) between-group differences (Healthy Control vs Insulin Resistant; HC vs IR) in resting-state functional connectivity measured by fMRI (n=18 HC, and n=56 IR), arranged by hemisphere and lobe. Stronger functional connection in HC (109 connections) and IR (49 connections) is shown as green and purple ribbons, respectively. Ribbon width corresponds to the magnitude of group differences in functional connectivity, computed as the Fisher z-transformed correlation difference (Δz) and weighted by statistical significance (αΔz). Occip. = Occipital, Pariet. = Parietal, SC=Subcortical, Ins. = Insula, and Cb = Cerebellum. (B) Time-series examples show positive (in-phase) and negative (out-of-phase) connectivity (correlations), and the strength of connectivity was determined by how far the Fisher z-transformed correlation coefficient was from 0. (C) Most of the 171 observed group differences in connectivity were in the positive direction, with stronger positive connectivity in HC (dark green) accounting for 55% of group differences. Stronger positive connectivity in IR (dark purple) accounted for 28% of group differences. (D) Stronger positive Frontal-Limbic and Frontal-Temporal connectivity was most prominent in the HC vs IR groups. (E) The stronger positive connections in the HC vs IR group were associated with connectivity between the Executive Control and Default Mode networks. (F) A functional connection hub map demonstrates the specific frontal, limbic, temporal, and parietal region-to-region connections with stronger positive relationships that accounted for the majority of the HC vs IR group differences. Line width and color demonstrate the strength of connection (based on Fisher’s Z Correlation Coefficients), text and node size increase as the number of connections increase. (G-H) Stronger positive connections in the IR vs HC groups were more diffusely distributed across lobes and networks.

Figure 3. Insulin Resistance is Related to Lower Brain Volume.

(A) Whole brain (excluding ventricles) (B) total cerebral white matter, (C) total grey matter, and (D) and hippocampal volumes (normalized to estimated intracranial volume, eTIV) in Healthy Controls (HC; n=17; green) and Insulin Resistant (IR; n=56; purple) participants are displayed as mean (bars) with participants represented as individual dots. (E) Regional white matter brain volumes that were significantly different (P<0.05) between HC and IR are displayed. Two-sided Student’s t-tests were used to compare group differences. (F-I) X-Y plots demonstrate the significant negative correlation between fasting glucose and total brain volume (F), total cerebral white matter volume (G), frontal pole white matter volume (H), and anterior corpus callosum volume (I). Dots represent individual participants.

Figure 4. Loss of Brain Insulin Receptor Impairs Brain Mitochondrial Function and Memory in Mice.

(A) Mice with floxed insulin receptor (InsR) driven by the Nestin promoter what were negative (Cre-) or positive carriers of the CreERT2 gene (NIRKO; Neuron Insulin Receptor Knockout) were injected with tamoxifen (to induce brain InsR knockout in the NIRKO group) and brains were harvested 4 weeks later for quantification of cerebral InsR mRNA by qPCR (n=6 per group). (B) Representative immunoblots of the insulin receptor protein across multiple tissues demonstrate that reduction in insulin receptor protein with tamoxifen in the NIRKO mice was specific to the brain. (C) InsR protein was quantified in the whole brain (n=6 mice per group). (D) Blood glucose values during a fasted (11hr overnight) intraperitoneal glucose tolerance test (ipGTT; 2g/kg of 20% D-glucose injected) and a fasted (6hr) intraperitoneal insulin tolerance test (ipITT; 0.75U/kg insulin injected) 4 weeks following tamoxifen treatment (n=10 per group). Lines graphs presented are presented as mean ± SD. (E) Cerebral mitochondrial DNA copy number (Nd1 and Nd4; n=6 per group) and citrate synthase activity (marker of mitochondrial abundance; n=8 per group). (F) Isolated cerebral mitochondrial ATP production rate (MAPR; n=9 per group) and cytochrome c oxidase (COX; complex IV) activity (marker of mitochondrial oxidative capacity n=8 per group). (G) Isolated cerebral mitochondrial reactive oxygen species (H₂O₂) production rate (n=9 per group) and superoxide dismutase 2 (SOD2) activity (marker of mitochondrial antioxidant defense; n=8 per group). (H) Working memory measured by 7-minutes spontaneous alternation behavior testing was quantified as % of maximum possible correct alternations (n=14 per group). (I) Mice were familiarized to a Y-maze for 5 minutes with one arm closed off, and then short-term memory was measured 3 hours later by assessing the time spent exploring the novel arm (now opened) during a 5-minute test (n=14 per group).

Figure 5. 40 Weeks of Metformin Improves Glucose Uptake in Specific Brain Regions.

(A) After passing inclusion and exclusion criteria, 40 participants with IR completed baseline testing (Cohort 2 from Supplemental Figure 1A) and were randomized in a double-blind fashion to 40-weeks of either 2000mg/day metformin or placebo. Baseline tests were repeated in all participants after 40-weeks of intervention. (B) The change (Δ) in in body mass index (BMI, n=20 per group), body composition (% fat, n=20 per group), HbA1c (n=20 per group), controlled fasting glucose (n=20 per group), and glucose and insulin area under the curve (AUC during oral mixed meal tolerance test, n=19 per group) after 40-weeks of either Placebo (Pla) or Metformin (Met) is shown as median (Black line) with participants represented as individual dots. (C) A statistical trend (P=0.086) towards group differences in the intervention-effect on whole brain glucose uptake, displayed as relative standardized uptake value (SUVr). (D) Standardized 3-dimensional FDG SSP (Stereotactic Surface Projection) intensity maps demonstrate Pre- and Post-Intervention (Placebo vs Metformin) glucose uptake across the whole brain. (E-H) Brain regions with statistically different intervention-induced change (Δ) in glucose uptake. Sample sizes were n=18 for Placebo and n=20 for Metformin for ¹⁸FDG-PET analysis. Two-way ANOVAs were used to compare group differences in Pre vs Post Intervention values and the p-value of the interaction effect is displayed.

Figure 6. 40 weeks of metformin enhances functional connectivity and prevents age-related brain atrophy.

(A) Chord diagram represents all significant (FDR<0.05) between-group differences (Metformin vs Placebo) in the 40-week change in resting-state functional connectivity measured by fMRI (n=20 per group), arranged by hemisphere and lobe. Greater change in functional connection in the metformin group (100 connections) and placebo group (25 connections) is shown as orange and teal ribbons, respectively. Ribbon width corresponds to the magnitude of group differences in functional connectivity, computed as the Fisher z-transformed correlation difference (Δz) and weighted by statistical significance (αΔz). Occip. = Occipital, Pariet. = Parietal, Temp. = Temporal, SC=Subcortical, Ins. = Insula, and Cb = Cerebellum. (B) Number of region-to-region connections (based on fMRI) that became more positively (bars going up, left panel) or less positively (bars going down, left panel) connected or more negatively (bars going up, right panel) or less negatively (bars going down, right panel) connected with either placebo (teal) or metformin (orange). (C) The functional connections that became more positive with metformin were included in bubble diagram demonstrating the frontal lobe was largely responsible for significant group differences altered lobe-to-lobe connectivity. (D) The functional connections that became more positive with placebo were included in bubble diagram demonstrating fewer positive connectivity differences observed in response to placebo than metformin. (E) The change (Δ) in whole brain (excluding ventricles) volume (normalized to estimated intracranial volume, eTIV) after 40-weeks of either Placebo (Pla) or Metformin (Met) is shown as means (bars) with participants represented as individual dots. (F-G) The change (Δ) in the white matter volumes of the Lateral Orbital Frontal Cortex and the Pars Opercularis after 40-weeks of intervention. Sample sizes were n=18 for Placebo and n=17 for Metformin for volumetric analysis. (H) The change in composite measures of Processing Speed and Working Memory subtests from the Patient-Reported Outcomes Measurement Information System (PROMIS) instruments in the NIH Toolbox-Cognition. The change in the T scores from baseline to 40-weeks was compared between placebo or metformin (n=20 per group). Data are presented as mean ± SD. Dots represent individual participants.