The independent effects of plasma glucose and insulin on brain function are not clearly defined. Chronically elevated plasma glucose is characteristic of type-1 diabetes (T1D). Altered brain function has been associated with poor glucose control in T1D^1,2, and hyperglycemia has been implicated in associated cognitive and mood impairments³. Elevated plasma insulin is often present in type-2 diabetes (T2D) and metabolic syndrome. These metabolic disorders have been associated with altered brain function⁴ and an elevated risk for mild cognitive impairment and Alzheimer’s disease⁵; chronic hyperinsulinemia related to insulin resistance may also be a contributing factor⁶. Yet, the mechanisms by which these metabolic disorders affect brain metabolism and function are still poorly understood. The goal of this study was to elucidate the independent effects of hyperglycemia and hyperinsulinemia on regional brain function measured by amplitude of low frequency fluctuations (ALFF) using resting state functional magnetic resonance imaging (rs-fMRI) in healthy individuals. We measured whole brain ALFF⁷ and fractional ALFF (fALFF)⁸ in slow-band 5 (SB5: 0.01-0.027 Hz) and slow-band 4 (SB4: 0.027-0.073 Hz)⁹ during 2 separate MRI scanning visits in 10 healthy non-diabetic subjects [mean (± SD) age = 28 ± 7 yrs, 6 M / 4 F, HbA1c = 5.5 ± 0.3%, fasting plasma glucose = 93 ± 6 mg/dl, fasting insulin = 4.0 ± 2.5 μU/ml]. Subjects were instructed to keep their eyes open and stare at a cross during a 6-minute rs-fMRI run. During visit 1, rs-fMRI was performed in the fasted euglycemic (EU1) state followed by a hyperglycemic (HG) clamp during which a primed-variable glucose infusion was used to raise plasma glucose concentration by approximately 100 mg/dl for 60 minutes (mean glucose = 205 ± 17 mg/dl, insulin response = 36 ± 24 μU/ml). During visit 2, at least 15 days later, subjects were again studied in the basal state (EU2) followed by a hyperinsulinemic euglycemic clamp (EU-HI) during which a primed-continuous insulin infusion was administered to match the insulin levels achieved during the HG clamp, and a variable glucose infusion was administered to maintain euglycemia (mean glucose = 96 ± 7 mg/dl, insulin = 32 ± 21 μU/ml). All MRI data was acquired on a 3T GE Signa HDxt scanner. A T1-weighted structural scan (MPRAGE) was acquired for each session to aid registration of the fMRI data to standard MNI-152 space. A BOLD-EPI sequence was used for rs-fMRI (TR/TE=3000/27 ms, flip=8º, voxel size = 3.75 x 3.75 x 4 mm,180 volumes). We analyzed rs-FMRI data using the FSL (fsl.fmri.ox.ac.uk) and DPARSF (rfmri.org) software packages to obtain SB4 and SB5 ALFF and fALFF on a voxel-wise basis for each subject during each condition. We performed paired comparisons of SB5 fALFF during 1) EU1 vs. HG (Visit 1), 2) EU2 vs. EU-HI (Visit 2), and 3) HG (Visit 1) vs. EU-HI (Visit 2) using the general linear model.SB5 fALFF was significantly decreased in the left medial frontal gyrus (MFG) and right posterior cingulate (PCC) and cuneus/precuneus cortices during HG relative to EU1 (p<0.01, corrected - Fig. 1), but no significant difference was found during EU-HI relative to EU2. No significant difference was found in these regions between the HG (hyperglycemia) and EU-HI (matched hyperinsulinemic euglycemia) clamps, but SB5 fALFF was decreased in the left putamen during HG relative to EU-HI (p<0.01, corrected - Fig. 2). Thus, high plasma glucose combined with high insulin significantly decreased MFG/PCC SB5 fALFF, while high plasma insulin alone with normal glucose levels had no significant independent effect on SB5 fALFF.

Our results demonstrate that acute high plasma glucose and insulin, such as may be encountered after a high glycemic load meal, decreases SB5 fALFF in the MFG and PCC/precuneus, while matching high plasma insulin has no significant independent effect on these brain regions in healthy individuals. ALFF and fALFF have been proposed to reflect the regional intensity of spontaneous neuronal activity (SNA)^10,11. We evaluated ALFF and fALFF as proposed by Zuo et al.¹² in distinct frequency bands based on work suggesting that independent frequency bands are generated by distinct oscillators⁹. We focused on fALFF because it was shown to be less sensitive to signal fluctuations due to physiological noise⁸, and on SB5 because it was shown to be preferentially localized in cortical grey matter¹². Our results suggest that SNA decreases in MFG and PCC/precuneus, regions associated with executive function and the default mode respectively, during hyperglycemia. These findings on the acute effects of hyperglycemia may be helpful in understanding brain functional adaptations to chronic hyperglycemia in diabetes, and their implications for comorbid neuropsychiatric complications.