mTOR inhibition enhances synaptic and mitochondrial function in Alzheimer’s disease in an APOE genotype-dependent manner

Abstract

Apolipoprotein ε4 (APOE4) carriers develop brain metabolic dysfunctions decades before the onset of Alzheimer’s disease (AD). A goal of the study is to identify if rapamycin, an inhibitor for the mammalian target of rapamycin (mTOR) inhibitor, would enhance synaptic and mitochondrial function in asymptomatic mice with human APOE4 gene (E4FAD) before they showed metabolic deficits. A second goal is to determine whether there may be genetic-dependent responses to rapamycin when compared to mice with human APOE3 alleles (E3FAD), a neutral AD genetic risk factor. We fed asymptomatic E4FAD and E3FAD mice with control or rapamycin diets for 16 weeks from starting from 3 months of age. Neuronal mitochondrial oxidative metabolism and excitatory neurotransmission rates were measured using in vivo ¹H-[¹³C] proton-observed carbon-edited magnetic resonance spectroscopy, and isolated mitochondrial bioenergetic measurements using Seahorse. We found that rapamycin enhanced neuronal mitochondrial function, glutamate-glutamine cycling, and TCA cycle rates in the asymptomatic E4FAD mice. In contrast, rapamycin enhances glycolysis, non-neuronal activities, and inhibitory neurotransmission of the E3FAD mice. These findings indicate that rapamycin might be able to mitigate the risk for AD by enhancing brain metabolic functions for cognitively intact APOE4 carriers, and the responses to rapamycin are varied by APOE genotypes. Consideration of precision medicine may be needed for future rapamycin therapeutics.

Introduction

Apolipoprotein ε4 (APOE4) allele is the most significant genetic risk factor for late-onset Alzheimer’s disease (AD) (>65 years). ¹ Individuals with one or two copies of APOE4 have 2- and 4-fold increased risk for AD, respectively, compared to non-carriers. ¹ Cognitively normal APOE4 carriers develop metabolic deficits decades before the aggregation of beta-amyloid (Aβ) and neurofibrillary tau tangles.^{2 –6} Specifically, lower glucose uptake, or hypometabolism, was shown in brain regions associated with AD in asymptomatic middle-aged adults carrying APOE4 alleles before AD neuropathology and clinical symptoms become apparent.^{2 –6} Clinical studies suggest that glucose hypometabolism is a better predictor for cognitive decline than Aβ and tau levels in patients with late mild cognitive impairment and AD.^{7 –10}

Early intervention to protect brain metabolic function may be critical to prevent AD onset or at least to slow its progression for APOE4 carriers. Rapamycin is an inhibitor for the mammalian target of rapamycin (mTOR), which is the serine/threonine kinase that regulates the response of eukaryote cells to nutrients, growth factors, and cellular energy status.^11,12 Inhibition of mTOR has been shown to extend longevity in various species, promote health span and reduce chronic disorders in mammals.^13,14

Recent studies showed that mTOR inhibitors preserve cerebrovascular, synaptic density, and cognitive functions in normative aging rats, ¹⁵ reduce Aβ and tau deposition, and improve memory in mice with symptomatic AD.^16,17 We also showed that rapamycin restores brain vascular function, clears cerebral amyloid angiopathy (CAA) and enhances blood brain barrier integrity in symptomatic AD mice and asymptotic APOE4 mice.^{18 –20} However, it is unknown whether rapamycin could impact mitochondrial integrity and synaptic transmission in healthy, asymptomatic APOE4 carriers.

In this study, we used an advanced APOE4 mouse model that mimics human AD (the E4FAD mice).^{21 –23} The healthy young E4FAD mice were fed with rapamycin before they started to develop Aβ and cognitive impairments. ²⁴ We used novel in vivo proton-observed carbon-edited (POCE; ¹H-[¹³C]) magnetic resonance spectroscopy (MRS) technique to measure mitochondrial oxidative metabolism and neurotransmission rates, as well as Seahorse for ex vivo mitochondrial respiration measurements. We performed the same measurements for mice with human APOE ε3 alleles (the E3FAD mice), a neutral AD genetic risk factor, to identify whether there are genetic-dependent responses to rapamycin. Our goal is to determine if early intervention with rapamycin could enhance brain mitochondrial integrity and energy metabolism in young E4FAD mice before they show metabolic deficits, and whether this intervention would be APOE genotype-dependent, which could potentially impact precision medicine approaches for future rapamycin therapeutics.

Materials and methods

All experimental procedures with mice were performed according to NIH guidelines and approved by the Institutional Animal Care and Use Committee (IACUC) at the University of Kentucky (UK) and Yale University (YU). ARRIVE guidelines were followed for preparation of the manuscript. We used a C57BL/6 mouse model which accumulates human Aβ42 due to co-expression of 5 familial-AD (5xFAD) mutations (APP K670N/M671L + I716V + V717I and PS1 M1461L + L286V) in conjunction with human targeted replacement APOE (ε4 in the E4FAD line and ε3 in the E3FAD line). These models were described in our previous studies. ²⁵ Both male and female mice (N = 40 in total) were used in the study, and were randomly assigned to the following groups: (i) mice fed with rapamycin supplemented diet: E3FAD-Rapa (n = 9; M:F = 5:4) and E4FAD-Rapa (n = 9; M: F = 5:4) or (ii) mice fed with control diet: E3FAD-Ctrl (n = 11; M:F = 7:4) and E4FAD-Ctrl (n = 11; M:F = 6:5). These diets started from 3 months of age and continued for 16 weeks based on our previous feeding protocol.^19,25 The sample size was determined via power analysis to ensure comparison at a 0.05 level of significance and 90% chance of detecting a true difference of each measured variable between the Rapamycin vs. Control groups. Rapamycin was mixed in the chow diet at a concentration of 14 parts per million (ppm). Given that the average mouse has a body weight of 30 grams and consumes 5 grams of food each day, this concentration results in an intake of approximately 2.24 mg of rapamycin per kilogram of mouse body weight per day. The detailed protocol of Rapamycin treatment was described in our previous studies. ²⁵ At the end of 16 weeks the animals were transferred from UK to YU for conducting the MRS experiments, which were performed after a quarantine of 2 weeks to ensure the health of the animals after transfer.

The mice were fasted over 12 hours with free access to water and anesthetized using urethane (1.5 g/Kg B.W). The mice were spontaneously breathing with a mixture of 30% O₂/70% N₂O under urethane anesthesia throughout the experiment. The respiration rate was continuously monitored (SA Instruments Inc., Stony Brook, NY, USA) and the core body temperature was maintained at ∼37°C using a water heating pad.

In vivo MRS measurements

We used the ¹H-[¹³C] or POCE-MRS technique, which enables dynamic detection of mitochondrial function and neurotransmission rate. ²⁶ The advantage of POCE over conventional ¹³C-MRS is its 4-fold enhanced sensitivity because it detects the ¹H signals directly bound to ¹³C nuclei, allowing measurement of ¹³C-enrichment of various metabolites (e.g., glutamate, glutamine or lactate). To measure the ¹³C turnover in real time we infused glucose labeled with ¹³C at positions C1 and C6. During [1,6-¹³C] glucose infusion, in the first pass of the tricarboxylic acid (TCA) cycle glutamate is labeled first in the C4 position followed by glutamine in the C4 position, whereas in the second pass of the TCA cycle glutamate/glutamine are labeled in the C2/C3 positions.

The in vivo POCE spectra were obtained on an 11.7 T BioSpec horizontal bore scanner (Bruker, Billerica, MA, USA) using a radio frequency (RF) surface coil (14 mm) tuned to proton frequency (499.8 MHz) positioned on top of the animal head. The RF decoupling on ¹³C (125.7 MHz) was done with two orthogonal, surface coils (21 mm) in quadrature mode covering both sides of the animal head. Static magnetic field (B₀) field homogeneity was optimized by adjustment of first- and second-order shims using B₀ field mapping followed by voxel shimming of water signal to achieve a linewidth at half height of less than 25 Hz. Localization was achieved with a LASER (localized adiabatic selective refocusing) pulse sequence. Spectra were collected with a repetition/echo time (TR/TE) of 4000/22.5 ms. The POCE spectra were collected from a localized voxel of 105 µL volume (5 × 3.5 × 6 mm³) encompassing cortical and hippocampal areas. The large volume was necessary to have sufficient sensitivity to obtain dynamic ¹³C turnover data. During the MRS scan, the mice received 0.24 mL of 0.75 mol/L [1,6-¹³C] glucose (99% enriched, Cambridge Isotopes, Andover, MA) which was infused through jugular vein under urethane anesthesia using a time-dependent infusion rate^27,28 that followed a decreasing exponential function during the first 8 minutes and was constant for the remaining of POCE data acquisition to raise the plasma glucose rapidly (<1 min) and maintain a nearly constant level and enrichment thereafter. The infusion rate was modified manually via a controlled pump system (Harvard Apparatus, Millis, MA). At the end of each experiment, the brains were frozen in situ with liquid nitrogen to instantly stop the metabolism as described previously. ²⁹ Brain extracts were prepared to obtain the end point of ¹³C enrichments of metabolites. During the in situ freezing the mice were spontaneously breathing under deep anesthesia to maintain blood oxygenation while liquid nitrogen was poured into a funnel sutured to the scalp. We used the in situ freezing method as described by Ponten U et al.. ³⁰ With in situ freezing the brain in the skull, freezes from the surface inward with blood flow (and oxygen) maintained ahead of the freezing front by continuous respiration, preventing hypoxia and maintaining metabolites near their in vivo levels. It takes a minute or so for the cortex to freeze and longer, for deeper structures. After funnel removal, the animal is decapitated, and the head frozen in liquid nitrogen and stirred at −80°C. The frozen brain tissue is then chipped out under liquid nitrogen irrigation.

Preparation of brain extracts

Brain extracts (of cerebrum only) were prepared to obtain the end point of ¹³C enrichments of metabolites. Metabolite concentrations were measured with ¹H-[¹³C] POCE NMR using a 500-MHz Bruker AVANCE vertical bore spectrometer (Bruker, Billerica, MA, USA). Ethanol extracts were prepared from the frozen brains (200 to 300 mg wet weight) where [2-¹³C]-glycine (50 μL, 5 mmol/L) was added as an internal concentration reference at the beginning of tissue extraction. After centrifugation the extract was suspended in 600 μL of a buffer solution containing 90% ethanol, 10% 0.1 M phosphate buffer, pH 7.4 followed by centrifugation for 30 min. Supernatant was removed, 900 µL of 60% ethanol and 40% water was added, and the samples were centrifuged again. This step was repeated two times. The collected supernatant was filtered through chelex-200 column and pH was adjusted. Samples were freeze dried using liquid nitrogen and lyophilized for 2 days. The extract powder buffer/D₂O (1:2) were added to total volume of 600 µL to prepare the NMR samples. The concentrations and ¹³C enrichments of glucose, glutamate (Glu) and glutamine (Gln) C4 and C3 resonance: Glu-C4, Gln-C4, Glu-C3 and Gln-C3 were measured with POCE under fully relaxed conditions (i.e., repetition time of 20 seconds).

Metabolic modelling

The time courses of in vivo amino acid ¹³C enrichments of Glu-C4, Glu-C3, Gln-C4, and Gln-C3 were fitted to a two-compartment (neuron–astrocyte) metabolic model using the plasma time courses of ¹³C-enriched glucose as the input. The steady-state levels of the amino acid pools from extracts were used to calibrate the in vivo amino acid time courses. The goal of the metabolic modeling was to derive the rates of the neuronal TCA cycle (V_TCA,N) and the glutamate–glutamine neurotransmitter cycling (V_cycle). ³¹ In this model, glutamate was assigned 10% to glia and 90% to neurons, whereas glutamine was assigned 90% to glia and 10% to neurons. The cell specific ratios of glutamate to glutamine were determined based on measurements by Van Den Berg & Garfinkel, ³² but also determined by early turnover times points in ¹³C MRS studies of humans ³³ and rats. ³⁴ The astrocytic TCA cycle flux (V_TCA,A) was set to 15% of total TCA cycle flux, the anaplerotic flux through pyruvate carboxylase (V_PC) was set to 20% of the rate of glutamine synthesis (V_gln), or the rate of glutamate–glutamine cycling (V_cycle). ³¹ The exclusive assignment of glutamine to glia was based on very low glutamine levels observed in neurons due to its immediate conversion to glutamate by the phosphate activated glutaminase as it enters the neurons.^31,35 The relationship between ¹³C enrichments of plasma glucose and brain metabolites is described by coupled differential equations with restrictions of mass and isotope balance (CWave 3.6 software (GF Mason, New Haven, CT, USA). Given that product (P) is made from substrate (S), the mass equations take the form:

$$\frac{d\left[P\right]}{dt}=\sum _{i=1}^m{\mathit{Vin}}_i-\sum _{i=1}^n{\mathit{Vout}}_i$$ (1a)

where [P] is the concentration of the product, while Vin_i and Vout_i are the respective rates of mass inflow and outflow, with m and n the individual inflows and outflows, respectively. The equations for the labeled product is similar, including ¹³C enrichment:

$$\frac{d\left[P^{\mathrm{*}}\right]}{dt}=\sum _{i=1}^m\frac{\left[S_i^{\mathrm{*}}\right]}{\left[S_i\right]}{\mathit{Vin}}_i-\frac{\left[P^{\mathrm{*}}\right]}{\left[P\right]}\sum _{i=1}^n{\mathit{Vout}}_i$$ (1b)

where the pools with asterisks represent the ¹³C labeled concentrations of S and P. The first term on the right represents the sum of all m individual isotopic flows into the product, each of them coming from a substrate S_i, for which the labeled fraction is [S_i*]/[S_i]. The second term represents the sum of all n individual isotopic flows out of the product, where [P*]/[P] is the labeled fraction of the product P. A simple model which describes the principles of mass and isotope balance equations was described by Fitzpatrick et al. ²⁷ and then later advanced by Mason et al, ³⁶ in which [1-¹³C] glucose was injected into rat using an infusion protocol approximated by a step function rising from natural abundance enrichment to 67% within ∼30 seconds. However, in our experiments we used [1,6-¹³C] glucose which doubles the ¹³C enrichment of all metabolites. The plasma glucose concentration and ¹³C enrichment was maintained constant for 60–90 min. During the injection of labeled glucose, MRS was used to detect the appearance of ¹³C in glutamate (and glutamine) in the brain. A simple but effective concept of modeling is shown if Vin_i and Vout_i are the same ³⁶

$$\frac{d\left[P^{\mathrm{*}}\right]}{dt}=\left(\frac{\left[S^{\mathrm{*}}\right]}{\left[S\right]}\right)V-\left(\frac{\left[P^{\mathrm{*}}\right]}{\left[P\right]}\right)V$$ (2a)

providing an exact solution to the differential equation:

$$\frac{{[P}^{\mathrm{*}}]}{\left[P\right]}=\frac{\left[S^{\mathrm{*}}\right]}{\left[S\right]}\mathrm{\hspace{0.17em}}[1-e^{-\lambda t}]$$ (2b)

where λ = V/[P]. However, it is more common for numerical analysis of equation (1b) to accurately determine the fluxes, where the differential equations are solved using a first-order Runge-Kutta algorithm and fitting optimization is achieved with the Levenberg–Marquardt algorithm. We employed this approach in the current work using Matlab 2020 b (Natick, MA, USA). The uncertainties in the absolute fluxes obtained by fitting the metabolic model to the time course data were assessed by Monte-Carlo simulation with 100 iterations using CWave 3.6 software.^34,37,38

Quantification of the ex vivo spectra from brain extracts

The ex vivo spectra were processed and analyzed using TopSpin Bruker 4.1 (Bruker, Billerica, MA, USA). Briefly, spectra were phased, and frequency shifted to adjusting the shape of the peaks so that they all have an absorptive, positive line shape with corrected baseline. Scans with edit ON and edit OFF scans were then aligned before subtraction. Then LCModel was used for analyzing the spectra simulated basis sets: one for standard ¹H metabolites which was used to fit the edit OFF spectra, and one for selected ¹³C enriched metabolites which was used to fit the POCE difference spectra. Concentration of each metabolite of interest was calculated by integration of the peak of interest divided by the number of protons it corresponds to.

Mitochondrial isolation

Separate cohorts of male and female mice (N = 25 in total) were used for the Seahorse mitochondrial isolation experiments at UK, with E3FAD-Rapa (n = 7; M: F = 4:3), E3FAD-Ctrl (n = 5; M: F = 3:2), E4FAD-Rapa (n = 7; M: F = 4:3), E4FAD-Ctrl (n = 6; M: F = 3:3). The mouse cortex and hippocampus tissues were collected in 2 ml mitochondrial isolation buffer (215 mM mannitol, 75 mM sucrose, 0.1% bovine serum albumin (BSA), 20 mM HEPES, 1 mM EGTA; pH adjusted to 7.2 with KOH) containing 1 mM EGTA. The mitochondrial isolation procedure, including reagents, followed the protocol described previously by our group.^39,40 The entire isolation procedure was carried out at 4°C until the Seahorse bioenergetics estimation. The tissues were homogenized using a Teflon glass homogenizer and were transferred to 2 ml centrifuge tubes. The homogenates were spun at 1300 g for 3 minutes. The pellet containing nuclei and other cellular debris was discarded and the mitochondrial rich supernatants were transferred in a new 2 ml centrifuge tubes and were spun at 13000 g for 10 minutes. The mitochondrial pellets were then resuspended in 450 µl isolation buffer and placed in a nitrogen cell disruptor (Parr Instrument Company). To release synaptosomal mitochondria, the suspensions were subjected to pressurized N₂ 1200 psi for 10 minutes at 4°C and the pressure rapidly released to burst synaptosomes. The suspensions containing total (synaptic and non-synaptic) mitochondria were transferred to 1.5 ml centrifuge tubes, topped with isolation buffer and centrifuged at 13,000 g for 10 minutes. The supernatants were discarded, and the pellets were resuspended in 50–100 µl isolation buffer at a concentration of >10 mg/ml till assessment. Mitochondrial protein was estimated using a standard BCA method (Pierce BCA Protein Assay Kit, Thermo Fisher, catalog number 23227).

Mitochondrial respiration measurements

The mitochondrial experiments were performed on the XFe96 Flux Analyzer (Agilent Technologies, USA) using differentially isolated cortical and hippocampal mitochondria.^39,41,42 Briefly, mitochondria (4 µg mitochondrial protein) were resuspended in 30 µl ice cold respiration buffer (125 mM KCl, 0.1% BSA, 20 mM HEPES, 2 mM MgCl₂, and 2.5 mM KH₂PO₄, adjusted to pH 7.2), loaded per well on a Seahorse culture plate and centrifuged at 3000 g for 6 minutes. The wells were topped with additional 145 µl of respiration buffer preincubated at 37°C. The injection port solutions were prepared in respiration buffer without BSA and each injection port was filled with 25 µl of the respective combination of substrate/uncoupler/inhibitor. Port A contained ADP, pyruvate and malate, port B Oligomycin, port C FCCP and port D Succinate and Rotenone. The mitochondria were exposed to the final concentration of the mitochondrial modulators after each injection at 5 mM pyruvate, 2.5 mM malate and 1 mM ADP (via Port A), 2.5 μM oligomycin A (via Port B), 4 μM FCCP (via Port C) and 1 μM rotenone and 10 mM of succinate (via Port D). After each injection, the mitochondrial oxygen consumption rates (OCR) were measured as coupled or uncoupled states of respiration as mentioned previously.^39,42

Statistical analysis

Statistical analyses were performed using Graph Pad Prism, JMP Pro16, and SPSS. The OCR values were used for analysis within a given experiment. A randomized block design with a blocking factor of assay day was used to account for the day-to-day variability between the Seahorse plates. Two-way ANOVA was performed to compare the control vs rapamycin diet and E3FAD vs E4FAD factors. When appropriate, post-hoc analyses were carried out using the Šidák test. For the ¹³C enrichment in the POCE and brain extract ¹H metabolite experiments, we selected a One-way ANOVA due to the unique nature of the data, including the power and variance among groups. This choice was informed by the primary factor under investigation, which was the response to Rapamycin treatment (e.g., E3FAD-Ctrl vs, E3FAD-Rapa; E4FAD-Ctrl vs, E4FAD-Rapa). Post hoc Tukey test was performed to identify the significant differences between groups.

Results

Rapamycin enhances in vivo brain metabolism in E4FAD mice

After the 16 weeks of feeding period, we performed the in vivo POCE experiments by infusing ¹³C-labeled glucose through the jugular vein of the mouse (Figure 1(a)) while imaging in a horizontal 11.7 T Bruker MRI scanner (Figure 1(b)). The data was acquired from a localized voxel (5 × 3.5 × 6 mm³) shown as a red rectangle (Figure 1(c)) and spectra at various times were acquired during infusion of [1,6-¹³C₂]-D-glucose over 80 min (Figure 1(d)). ¹³C-labeled ¹H signals from the [4-¹³C]-Glutamate (Glu-C4) at ∼2.35 ppm and [4-¹³C]-Glutamine (Gln-C4) at ∼2.45 ppm, as well as their [3-¹³C]-Glu/Gln (Glx3) at ∼2.1 ppm can be observed over the course of the acquisition session (Figure 1(d)). Due to spectral overlap the Glu-C3 and Gln-C3 peaks could not be measured separately. Similarly, the signal intensity of Glx-C2 at ∼3.8 ppm increases over time. Figure 1(e) shows the dynamic time courses of ¹³C enrichments of Glu-C4 and Gln-C4, and best fits of the metabolic model for group averaged of E3FAD mice with the control diet (E3FAD-Ctrl) and the rapamycin diet (E3FAD-Rapa). Similarly, Figure 1(f) shows the dynamic time course for the E4FAD mice with the control diet (E4FAD-Ctrl) and the rapamycin diet (E4FAD-Rapa). The quantitative data are shown in Table 1. Additional details for 10-, 40- and 80-minutes time-points are provided in Supplementary Table 1 (Table S1). ¹³C labeling of Glu-C4 and Gln-C4 appeared earlier in both E3FAD-Rapa and E4FAD-Rapa, indicating that ¹³C-glucose was mainly oxidized in the neuronal compartment. E3FAD-Rapa mice achieved a 40% fractional enrichment in ¹³C Glu-C4 labeling by 40 min whereas the E3FAD-Ctrl only reaches the same level of fractional enrichment by 80 min. Similarly, the labeling curve for Gln-C4 reaches 30% fractional enrichment by 45 min for E3FAD-Rapa but E3FAD-Ctrl reaches the same level at 80 min. Labeling of Glx-C3 E3-Rapa and E3-Ctrl showed similar enrichment curves. For E4FAD only Gln-C4 signal was statistically higher in the E4FAD-Rapa group compared with their controls, suggesting that rapamycin was able to significantly enhance mitochondrial oxidative metabolism in neurons.

Figure 1.

POCE MRS in vivo and kinetics of ¹³C turnover into glutamate and glutamine pools from glucose. Following (a) ¹³C glucose infusion, (b) mice were scanned in the MRI scanner. (c) Voxel size and position (105 µL red square) in mouse brain. (d) Sample of time resolved ¹H[¹³C] or POCE spectra acquired from a voxel during infusion of [1,6-¹³C₂]-glucose over 80 min at 11.7T. Metabolic turnover of ¹³C label is observed in the ¹H signals from Glu4 at ∼2.35 ppm and Gln4 at ∼2.45 ppm, and their Glx3 and Glx2 counterparts at ∼2.1 ppm and ∼3.8 ppm, respectively. Time courses of ¹³C fractional enrichment (FE, %) on the left Y-axis of Glu4, Gln4, and GlxC3 pools (dots) during [1,6-¹³C₂]-glucose infusion and best fits of the metabolic model for Control (blue) and Rapamycin (tan) groups. Lines represents best fit of the constrained, two-compartment metabolic model to Glu4, Gln4 and Glx3 enrichment time courses, providing flux estimates for the localized brain region. for E3FAD Control and Rapamycin groups (e) and E4FAD Control and Rapamycin groups (f). Statistical significance of between group differences in ¹³C enrichments at the different time points was assessed by One-Way ANOVA and Tuckey’s post-hoc test.

Table 1.

Fractional enrichment (FE (%)) ¹³C-labeled metabolites at the end of ¹³C-labeled glucose infusion.

FE (%)	E3FAD-Ctrl(n = 5)	E3FAD-Rapa(n = 5)	p	E4FAD-Ctrl(n = 11)	E4FAD-Rapa(n = 7)	p
Glu-H4	39.4 ± 3.7%	43.9 ± 4.1%	0.07	43.1 ± 3.9%	48.8 ± 2.2%	0.05
Gln-H4	32.2 ± 3.1%	36.8 ± 5.2%	0.08	36.3 ± 1.8%	45.4 ± 2.7%	0.04**
Glx-H3	26.6 ± 2.4%	28.7 ± 2.9%	0.06	25.9 ± 4.0%	30.3 ± 3.9%	0.06

^**p < 0.05.

Rapamycin enhances neuronal mitochondrial activity and excitatory neurotransmitters in E4FAD mice

Using the in vivo POCE data, we were able to calculate the total glutamate–glutamine neurotransmitter cycling (V_cycle) and total neuronal TCA cycle (V_TCA,N) as illustrated in Figure 2(a), where neuronal glucose oxidation (CMR_glc(ox),N) is half of V_TCA,N. The fluxes for V_cycle in E3FAD-Rapa and E4FAD-Rapa were significantly increased compared to their control groups (Figure 2(b)). Only E4FAD-Rapa group shows a significant increase in V_TCA,N compared to their controls (Figure 2(c)). To better compare the results across various regions, groups and species, we used the ratio of V_cycle to CMR_glc(ox),N as an index. CMR_glc(ox),N is the specific glucose oxidation in neurons, which is approximately half of the V_TCA,N value. A significant increase (p < 0.05) in V_cycle/CMR_glc(ox),N was observed for the E4FAD-Rapa vs E4FAD-Ctrl group (Figure 2(d)). The results indicate that rapamycin enhances synaptic and mitochondrial activities in the young healthy E4FAD mice.

Figure 2.

Quantitative fluxes derived from in vivo POCE for E3FAD and E4FAD mice on rapamycin diet. (a) Illustration of resting total neurotransmitter cycling (V_cycle), total neuronal TCA cycle (V_TCA,N), and neuronal glucose oxidation (CMR_glc(ox),N). (b) Rapamycin significantly increased V_cycle in both E3FAD (p = 0.045) and E4FAD (p = 0.032) mice (i.e., 0.13 ± 0.03 vs. 0.20 ± 0.02 µmol/g/min for E3FAD-ctrl vs. E3FAD-rapa; and 0.12 ± 0.03 vs. 0.21 ± 0.03 µmol/g/min for E4FAD-ctrl vs. E4FAD-rapa). (c) Rapamycin did not significantly increase V_TCA,N in E3FAD mice (i.e., 0.23 ± 0.02 vs. 0.30 ± 0.03 µmol/g/min for E3FAD-ctrl vs. E3FAD-rapa). However, in E4FAD mice Rapamycin significantly increased V_TCA,N (i.e., p = 0.042; 0.26 ± 0.03 vs. 0.37 ± 0.04 µmol/g/min for E4FAD-ctrl vs. E4FAD-rapa). (d) Comparison of V_cycle/CMR_glc(ox),N ratio from current results (blue and red bars for E3FAD and E4FAD mice). Rapamycin significantly increased V_cycle/CMR_glc(ox),N ratio in E4FAD (*p < 0.05), but not E3FAD. The error bars represent the standard deviation.

Rapamycin increases glycolysis and inhibitory neurotransmitters in the E3FAD mice

At the end of each in vivo POCE experiment, the brains were frozen under liquid nitrogen and the brain tissue was extracted. Experimental setup for the ex vivo POCE at 11.7 T from brain extracts are shown in Figure 3(a). Differences between ¹H spectrum acquired with and without a ¹³C inversion pulse from brain extracts are shown in Figure 3(b). The difference spectrum shows the ¹H signals bound to ¹³C atoms of Glu, Gln, combined Glu-Gln pools (Glx), lactate (Lac), Alanine (Ala). The spectra with (blue) and without (red) ¹³C inversion pulses are demonstrated in the Supplementary Figure 1 (Fig. S1). The GABA level was measured at 1.9 ppm (C3). We used the LCModel ⁴³ to estimate the concentration of metabolites in E3FAD/E4FAD control and rapamycin groups. Lac, Ala, and GABA were found to be significantly different (p < 0.05) between E3FAD-Ctrl and E3FAD-Rapa (Figure 3(c)). Other metabolites showed an increasing trend in the E3FAD-Rapa group, but they did not pass the significance test.

Figure 3.

Effect of rapamycin diet on cerebrum metabolites measured by ex vivo ¹H MRS. (a) The ex vivo ¹H-[¹³C] spectra of brain tissue extracts acquired at 11.7 T were used to obtain the end point of ¹³C metabolites enrichments after 80 min of ¹³C-labeled glucose infusion. (b) POCE data acquired from brain extracts. The difference spectrum shows the ¹H signals bound to a ¹³C nucleus, which only occurs for the 1.1% natural abundance of Glu-H4 (2.35 ppm) and GlnH4 (2.45), Glx-H3 (2.1 ppm), Lac-H3 (1.3 ppm), Ala-H3 (1.5 ppm) and GABA-H3 (1.9 ppm) for E3FAD and E4FAD groups. (c) LCModel estimated concentration values of all animals in each cohort for metabolites of interest are shown. Three metabolites, lactate, alanine and GABA were found to be significantly different between E3FAD cohort. Statistical significance was not observed for E4FAD-rapa and E4FAD-ctrl cohorts using MANOVA with post-hoc Tukey test *p < 0.05. The error bars represent the standard deviation.

Rapamycin increases the overall mitochondrial respiration in the E3FAD mice

Mitochondrial function was assessed in isolated mitochondria from cortical and hippocampal regions. The oxygen consumption rates for all cell types in the sampled tissues were determined using the Seahorse platform (Figure 4(a)). The rates of respiration were determined as ATP synthesis rate driven by complex I as State III respiration, proton leak after inhibiting the ATPase complex using oligomycin as state IV respiration, uncoupling of the electron transport chain from the ATPase using protonophore FCCP as state V (CI) and inhibition of complex I using rotenone and addition of complex II substrate succinate to drive State V (CII) respiration (Figure 4(b)). From the bioenergetics determined in the cortical region, there was a significant improvement in E3FAD-Rapa mice across all the states of respiration except State IV (Figure 4(c)). Similarly, in hippocampus, E3FAD-Rapa mice showed significantly increased bioenergetics in the State III and state V(CII) states of respiration (Figure 4(d)). In contrast, these bioenergetic increases were not found in the E4FAD-Rapa, compared to their respective controls. The results indicate that when considering all cell types (and not just neurons), rapamycin was able to increase the overall mitochondrial respiration of the E3FAD, but not the E4FAD mice. Details of the statistical analysis are shown in the Supplementary Table 2 (Table S2).

Figure 4.

Effect of rapamycin diet on cortex and hippocampal mitochondrial bioenergetics by ex vivo Seahorse assay. (a) Isolation of mitochondria from cortical and hippocampal tissues. (b) Types of measurements include State III – ATP linked respiration, State IV – proton leak, State V (CI) – complex I mediated uncoupled respiration, and State V (CII) – complex II mediated uncoupled respiration (or maximal respiration). (c) Cortex mitochondrial bioenergetics. (d) Hippocampal mitochondrial bioenergetics. The data is shown as mean ± SEM, N = 5–7, *p < 0.05; **p < 0.01; ***p < 0.001.

Discussion

Using in vivo POCE, we demonstrated that rapamycin enhances mitochondrial function of neurons, glutamate-glutamine cycling, and neuronal TCA cycle rates in asymptomatic E4FAD mice. This shows that rapamycin can protect neuronal metabolic functions and levels of excitatory neurotransmitters before the development of AD symptoms for APOE4 carriers. These findings have significant implications because glucose hypometabolism has been considered a key indicator for cognitive impairment. Brain metabolic deficits are more closely linked with cognitive decline than Aβ and tau levels in patients with late mild cognitive impairment and AD.^{7 –10} Other studies using machine learning also demonstrate that glucose metabolic dysfunction can be a predictor for AD with an average of 75.8 months prior to its final diagnosis with high accuracy. ⁴⁴ In a large clinical data analysis from the Framingham Heart Study, increased cerebral glucose levels in midlife are shown to be highly associated with the risk of AD. ⁴⁵ Alzheimer’s disease has been considered as the Type 3 diabetes evident by the glucose dysfunction and insulin resistance in the brain,^46,47 which has a tremendous impact on lipotoxicity as well as chronic inflammation. ⁴⁸ Furthermore, insulin regulates distinct pathways in the hypothalamus, hippocampus, and nucleus accumbens. Insulin shows its most robust effect in the hypothalamus and regulates multiple genes involved in neurotransmission to differentially modulate glutamate receptors; while suppressing multiple neuropeptides. Dysregulation of these pathways may have a role in causing the characteristic cerebral alterations of AD, upregulation of Aβ aggregation, tau hyper-phosphorylation, inflammation, oxidative stress, and mitochondrial dysfunction.^48,49

Interventions that can mitigate insulin resistance at an early stage could be critical to protect the brain against AD. Intranasal insulin administration has been used as one promising method, however their effectiveness remains controversial.^50,51 mTOR inhibition has been shown to be able to increase insulin sensitivity, ⁵² reduce inflammation^53,54 and extend longevity.^13,55 This is in line with our previous findings that rapamycin can significantly reduce Aβ retention, decrease lipotoxicity, restore free fatty acid level, and enhance recognition memory in asymptomatic E4FAD mice. ²⁵ As glucose uptake and neuronal activity are highly coupled with cerebral blood flow (CBF), ⁵⁶ the findings from the present study are also consistent with our previous report that rapamycin restores CBF (especially in females), blood brain barrier (BBB) activity for Aβ transport and neuronal integrity. ²⁵ It is also consistent with reports with other interventions, such as caloric restriction and ketogenic diet, that mTOR inhibition can enhance metabolic and vascular functions, reduce Aβ, slow down brain aging and mitigate risk for AD.^{26,57 –61} Our findings also align with previous in vivo research indicating that rapamycin offers neuroprotection without neurological side effects.^62,63 However, we recognize that certain in vitro studies have observed neurotoxic effects associated with rapamycin.^64,65 The inconsistencies between these outcomes may be attributable to differences in study design, such as the conditions of cell lines used in the in vitro studies, duration of rapamycin exposure and the dosages administered. It may be crucial to consider these variables when interpreting the results and their implications for future clinical applications.

Another major finding is that rapamycin has differential effects on the mice based on their APOE variance. It is suggested that rapamycin had more impact on neuronal mitochondrial oxidation and excitatory neurotransmission in the E4FAD mice, whereas in the E3FAD mice the impact was more on glycolysis, non-neuronal mitochondrial metabolism, and inhibitory neurotransmission. E3FAD-Rapa mice, compared to their controls, had higher levels of glycolysis and inhibitory neurotransmitters, and the overall, non-neuronal mitochondrial respiration in cortex and hippocampus. These patterns were not observed in the E4FAD-Rapa mice compared with their controls. In particular, rapamycin increases the expression of electron transfer proteins of the oxidative phosphorylation (OXPHOS) in the cortex and hippocampus of the E3FAD mice, especially components of the complex I–III–IV–V. This may result in increases of overall energy metabolism, a more efficient mitochondrial OXPHOS system, reduced reactive oxygen species in those brain regions related to cognitive functions. It implies that rapamycin might enhance non-neuronal, such as astrocytic functions, for the E3FAD mice. It is well documented that bioenergetics of astrocytes are predominantly driven by glycolysis to support neuronal activity, and may also play a critical role in AD development. ⁶⁶ These data suggest that E3FAD might have a better function in astrocytes induced by rapamycin, which may help reduce the risk of AD as well. However, future POCE studies with ¹³C-labelled acetate may be needed to study rapamycin effects on astrocyte metabolism. ¹³C-labeled acetate is a valuable tool for studying astrocytic metabolism because acetate can be selectively taken up by astrocytes.^67,68 This is primarily due to the expression of specific enzymes in astrocytes that are adept at metabolizing acetate. When acetate is labeled with ¹³C, it becomes a tracer that can be tracked to observe astrocytic metabolism, including fluxes of TCA cycle and neurotransmission. Since glutamate and GABA comprise the majority (∼90%) of synapses in the cerebral cortex, metabolic flux measured using ¹³C-labeled glucose and acetate permits a comprehensive assessment of neuronal and glial metabolism involved with glutamatergic and GABAergic neurons. ⁶⁹

Findings from the current study are also consistent with our previous report which suggests that there are APOE-dependent pharmacogenetic responses to rapamycin on vascular and cognitive functions. ²⁵ As aforementioned, rapamycin restored CBF, BBB activity for Aβ transport, neurotransmitter levels, neuronal integrity, and free fatty acid level, and reduced Aβ retention in the E4FAD mice, which were not observed in the E3FAD-Rapa mice. In contrast, E3FAD-Rapa mice had lower cerebrovascular reactivity responses and lower anxiety, which were not apparent in the E4FAD-Rapa mice. ²⁵ The different metabolic responses induced by rapamycin emphasize the importance of considering an individual’s genetic profile when considering disease treatment. The findings are also consistent with our report that AD patients with APOE3 and APOE4 alleles have different underlying metabolic pathway changes. ⁷⁰ Particularly, we showed that APOE4 carriers have deficits in metabolites associated with the TCA cycle and oxidative phosphorylation that were prominent in advanced stages compared to the early stages. Conversely, those with the APOE3 allele exhibit metabolite deficits linked to oxidative DNA damage and a reduction in inhibitory neurotransmitters like GABA, again more noticeable in later stages. These patterns suggest essential metabolic distinctions between APOE3 and APOE4 allele carriers that could influence the development of AD. Recent research has also pointed to potential differences in the microbiome based on APOE allele variation. Individuals with the APOE4 allele experience changes in gut microbiome diversity and composition, a condition known as dysbiosis, well before AD symptoms manifest.^71,72 This alteration in microbiome diversity and composition associated with APOE variation may also alter how the body metabolizes food and responds to dietary interventions,^{73 –75} which corresponds with our observations in this present study using rapamycin. Future research will be needed to investigate the mechanisms causing these varied responses to rapamycin between APOE3 and APOE4 carriers. Additionally, tailoring treatment based on pharmacogenetic and nutrigenetic profiles will be crucial for the advancement of personalized medicine in the future.

A unique aspect of the present study is the in vivo measurements with ¹³C-labeled glucose using the POCE method. The POCE techniques have only been used for human and rat studies due to resolution limitations and the level of ¹³C concentration. Previous studies in mice were exclusively ex vivo.^68,76 Here we show, for the first time, the use of in vivo POCE method in mice by overcoming limitations. POCE is a non-invasive neuroimaging method for the tracking and quantifying metabolic biomarkers, specifically neurotransmitter turnover and oxidative metabolism reflecting molecular processes in brain through ¹³C-labeled glucose.^26,29,56,77 POCE can reflect molecular processes in brain through ¹³C-labeled substrates like glucose and/or acetate. ⁷⁸ Since the natural abundance of the ¹³C isotope is only 1.1%, using ¹³C-labelled substrates allows measurement of the kinetics of ¹³C incorporation into cerebral metabolites. Turnover of ¹³C into various metabolite pools allows determination of metabolic rates, e.g., neuronal glucose oxidation (CMR_glc(ox),N) and neurotransmitter cycling (V_cycle). Majority of in vivo POCE studies in normal brain of humans and rodents show that the ratio of V_cycle/CMR_glc(ox),N is approximately 1. ⁷⁹ It is important to note that with rapamycin treatment the V_cycle/CMR_glc(ox),N ratio for E4FAD mice approaches 1, which suggests semblance of normal mitochondrial and synaptic health. The success of POCE presented in the current study will allow in the future many studies using mouse models because it will provide a measure of bioenergetics in different cell types, such as astrocytic activities through ¹³C-labeled acetate as described previously.

We have used MRI, MRS and PET neuroimaging to non-invasively detect the metabolic and vascular changes induced by mTOR inhibition.^{18 –20,25,61,80} In this study, we further included the POCE method to determine mitochondrial dynamics and synaptic activity in vivo. In the future, other imaging methods like calibrated fMRI to measure cerebral metabolic rate of oxygen consumption at high spatial resolution^{81 –83} with superior spatial resolution in mouse brain, ⁸⁴ can also be integrated to obtain a more comprehensive picture the effects of mTOR on brain metabolic functions.

A limitation of the study is that we did not have enough statistical power to dissociate sex-specific effects due the unequal number of male and female subjects during the POCE experiments, attributable to the uncontrollable factors inherent in breeding. We grouped both males and female subjects, which prohibits us to stratify sex effects. In our future studies, we will further determine the sex-dependent responses to rapamycin on brain metabolism.

In summary, our findings indicate that rapamycin might be able to mitigate the risk for AD by enhancing brain metabolic functions for cognitively intact APOE4 carriers, while enhancing astrocytic activities in the APOE3 carriers. Future studies may further determine whether rapamycin is also effective to restore brain metabolic functions in the APOE4 carriers after the deficits have occurred. Given that the responses to rapamycin are APOE genotype-dependent, precision medicine will need to be considered for future rapamycin therapeutics. As rapamycin is FDA-approved and POCE has been used in humans, the outcomes from the study are readily applicable to humans to determine whether rapamycin could mitigate brain metabolic deficits and thus AD risk for cognitively normal APOE4 carriers.

Supplementary material

Supplemental material for this article is available online.