Synaptic mitochondria glycation contributes to mitochondrial stress and cognitive dysfunction

Abstract

Mitochondrial and synaptic dysfunction are pathological features of brain ageing and cognitive decline. Synaptic mitochondria are vital for meeting the high energy demands of synaptic transmission. However, little is known about the link between age-related metabolic changes and the integrity of synaptic mitochondria.

To this end, we investigated the mechanisms of advanced glycation end product (AGE)-mediated mitochondrial and synaptic stress and evaluated the strategies to eliminate these toxic metabolites. Using aged brain and novel transgenic mice overexpressing neuronal glyoxalase 1 (GLO1), we comprehensively analysed alterations in accumulation/build-up of AGEs and related metabolites in synaptic mitochondria and the association of AGE levels with mitochondrial function.

We demonstrated for the first time that synaptic mitochondria are an early and major target of AGEs and the related toxic metabolite methylglyoxal (MG), a precursor of AGEs. MG/AGE-insulted synaptic mitochondria exhibit deterioration of mitochondrial and synaptic function. Such accumulation of MG/AGEs positively correlated with mitochondrial perturbation and oxidative stress in ageing brain. Importantly, clearance of AGE-related metabolites by enhancing neuronal GLO1, a key enzyme for detoxification of AGEs, reduces synaptic mitochondrial AGE accumulation and improves mitochondrial and cognitive function in ageing and AGE-challenged mice. Furthermore, we evaluated the direct effect of AGEs on synaptic function in hippocampal neurons in live brain slices as an ex vivo model and in vitro cultured hippocampal neurons by recording long-term potentiation (LTP) and measuring spontaneously occurring miniature excitatory postsynaptic currents (mEPSCs). Neuronal GLO1 rescues deficits in AGE-induced synaptic plasticity and transmission by full recovery of decline in LTP or frequency of mEPSC.

These studies explored crosstalk between synaptic mitochondrial dysfunction and age-related metabolic changes relevant to brain ageing and cognitive decline. Synaptic mitochondria are particularly susceptible to AGE-induced damage, highlighting the central importance of synaptic mitochondrial dysfunction in synaptic degeneration in age-related cognitive decline. Thus, augmenting GLO1 function to scavenge toxic metabolites represents a therapeutic approach to reduce age-related AGE accumulation and improve mitochondrial function and learning and memory.

Samanta et al. report that the accumulation of advanced glycation endproducts in synaptic mitochondria with ageing contributes to mitochondrial and synaptic stress, and to cognitive impairment. Removal or detoxification of glycation by enhancing the activity of the enzyme glyoxalase 1 restores neuronal and cognitive function in mice.

Introduction

Ageing is a high-risk factor for neurodegeneration and cognitive decline. Changes in metabolism and mitochondrial function have been identified as hallmarks of the ageing process, contributing to neurodegeneration and cognitive impairment.¹ Age-related metabolic changes lead to the accumulation of toxic metabolites such as advanced glycation end products (AGEs) and highly reactive glucose-derived methylglyoxal (MG), a precursor of AGEs. AGEs are post-translational protein modifications caused by the non-enzymatic binding of free amine groups and carbonyl groups of reducing sugars. It is now well accepted that the formation of AGEs is mainly caused by the action of various reactive metabolites other than glucose such as MG and MG-derived AGEs.^2-4 MG is the most potent precursor of AGEs in the rapid generation of glycated adducts on cellular and short-lived extracellular proteins, lipids and DNA.⁵

Increased levels of AGEs and their precursors in the brain are associated with ageing and various diseases, including diabetes, atherosclerosis, Alzheimer’s disease (AD) and cognitive decline.^6-10 AGE formation leads to the cross-linking of proteins and subsequent changes in the physicochemical properties of tissues, disrupting neuronal function. Thus, it is important to define the mechanisms underlying AGE-mediated mitochondrial and synaptic stress and the strategy to eliminate these toxic metabolites.

Glyoxalase 1 (GLO1) is the initial enzyme of the glyoxalase system, which detoxifies MG via the hemithioacetal, which is converted to S-D-lactoylglutathione and then hydrolyzed by GLO2 to D-lactate. Within this context, GLO1 represents the rate-limiting step for the detoxification of MG and related reactive dicarbonyls to prevent the formation and accumulation of AGEs. MG is a highly reactive metabolic intermediate of cellular metabolism through non-enzymatic degradation of the glycolysis-derived triose phosphates glyceraldehyde 3-phosphate and dihydroxyacetone phosphate. MG reacts with arginine, lysine and cysteine residues of proteins to form AGEs. MG is among the best-known glycation agents linked to age-related cell injury and cognitive impairment via their direct cytotoxicity and/or initiating the formation of MG-derived AGEs.^7,11-19 MG is also elevated in normal individuals who consume AGE-rich diets, and MG-AGEs in food are pathogenic.¹⁸ Increased levels of MG/AGEs are linked to cognitive dysfunction in ageing and AD.^4,20 Accordingly, GLO1 activity decreases with advancing age and progression of AD,^21,22 which could contribute to elevated levels of MG/AGEs. It is therefore of the highest interest and an attractive therapeutic target to eliminate harmful MG/AGEs. Strategies that reduce MG/AGEs levels in mitochondria and the brain by increasing GLO1 expression and activity are critical to consider as new avenues for both preventing and halting age-related dementia and cognitive decline.

Neurons are distinguished from many other eukaryotic cells by the long processes stemming from the cell body. Neuronal mitochondria are classified as non-synaptic (located in the cell body) or synaptic (located in the terminal part of cellular processes). Synaptic mitochondrial function is fundamental for neurotransmission and hippocampal memory formation. Because of intense energy demands and limited glycolytic capacity in neurons, synapses have the highest demand for mitochondrial ATP production.^23-25 Synaptic mitochondria are consequently more susceptible to oxidative stress and toxic metabolites, such as amyloid protein and abnormal tau in ageing brain and AD.^26-29 Thus, synaptic mitochondrial abnormalities may be important mechanisms underlying synaptic injury in age-related dementia and AD.

To date, little is known about the link between age-related metabolic changes and synaptic mitochondria. It is unclear whether AGEs accumulate predominantly in synaptic mitochondria, whether AGE-rich synaptic mitochondria are more vulnerable and whether clearance and detoxification of toxic metabolites (MG/AGEs) can reverse mitochondrial and synaptic dysfunction, improving learning and memory.

Herein, using aged brain and newly generated transgenic (Tg) GLO1 mice, we comprehensively analysed (i) the synaptic mitochondrial pool of AGEs and their relevance to mitochondrial function; (ii) the direct effects of AGEs on mitochondrial bioenergetics and synaptic plasticity; and (iii) the effect of exogenous AGEs on cognitive function in AGE-challenged mice. Finally, we investigated whether an increase in neuronal GLO1, a key enzyme for detoxifying MG, would reduce AGE accumulation and restore mitochondrial and cognitive functions in the AGE-enriched brain of a GLO1 mouse model. Additionally, AGE-insulted live brain slices were used as an ex vivo model to determine the direct effects of AGEs on mitochondrial and synaptic functions in live hippocampal neurons. These studies address the following unexplored key questions: Are synaptic mitochondria the focal point for AGE accumulation? Is AGE accumulation relevant to mitochondrial dysfunction? Does clearance/detoxification of AGEs following the gain-of-neuronal GLO1 function reduce synaptic mitochondrial AGE accumulation and restore mitochondrial and cognitive function? Do GLO1-producing neurons protect against AGE-induced synaptic perturbation? Our findings uncover a link between synaptic mitochondrial dysfunction and AGE metabolism in age-related synaptic failure and cognitive decline.

Materials and methods

Animal study

Animal studies were approved by the Institutional Animal Care and Use Committees of Columbia University and the University of Kansas. In this study, 3–30-month-old male and female wild-type (WT, C57BL/6) mice from Charles River and transgenic mice with neuronal overexpression of glyoxalase 1 (GLO1, Tg) generated in our laboratory were used. Mice were decapitated, and the cerebral cortex was used to isolate synaptic and non-synaptic mitochondria, which were then stored at −80°C for subsequent analysis, as described later.

Advanced glycation end product preparation and identification

AGE-bovine serum albumin (BSA) or MG-derived AGEs were prepared as previously described.^18,19 The degree of glycation was determined by immunoblotting using an anti-AGE antibody (generated in our laboratory) as previously described.¹⁶

Advanced glycation end product diet

Mice were fed with an AGE-rich diet (1.33% MG-derived AGEs) for 12 months from 9–21 months of age, as described in our previous study.¹⁸

Synaptic mitochondria preparation

For further details, see the Supplementary material.

Detection of methylglyoxal and advanced glycation end products

Quantification and analysis of MG in the isolated pure synaptic or non-synaptic mitochondria from brain tissue were performed using O-phenylenediamine (OPD) dihydrochloride derivatization by measuring quinoxaline under reverse phase high-performance liquid chromatography (RP-HPLC) as previously described.¹

Immunodot blot or ELISA to measure advanced glycation end products

Proteins (2 μg in each: confirmed by HSP60) were dot-blotted in triplicate on a nitrocellulose membrane, and the incorporated carboxymethyl-lysine (CML) and MG-AGE was immunochemically visualized and analysed by densitometry. A bar graph represents the quantification of the immunoreactive dot for CML and MG-AGE normalized to HSP60 using ImageJ (National Institutes of Health). The representative immunoblots show immunoreactive dots for MG-AGE (Cat. No. STA 011, Cell Biolabs) and CML (Cat. No. KAL-KH024, Cosmo Bio) from indicated synaptic and non-synaptic mitochondrial homogenates, and HSP60 (Cat. No. ADI-SPA-806-F, Enzo Life Sciences) served as a loading control. Levels of total AGEs, MG-AGE and CML were determined by ELISA using commercial ELISA kits (Cat. No. MBS704846, MBS756134 and MBS265592, MyBioSource, respectively).

Immunogold electron microscope

Mice were fixed by through-heart perfusion with cold 3% paraformaldehyde and 0.25% glutaraldehyde in PBS. Small pieces of cortex and hippocampus were additionally fixed overnight in the same fixative and embedded in LR White Resin. Ultrathin sections (∼400 nm) were blocked by 10% donkey serum (30 min, room temperature). Primary anti-AGE antibodies produced in our laboratory¹⁵ were applied at 1:50 overnight. Secondary antibodies conjugated with 12 nm gold particles were applied for 1.5 h at room temperature. Sections without AGE antibody were used as controls. Sections were contrasted with uranyl acetate and examined under an electron microscope (Jeol).

Mitochondrial respiration complex activity and ATP measurement

Mitochondrial respiration complex activity and ATP level were measured using homogenates of the cerebral cortex including hippocampus as previously described.³⁰

Mitochondrial respiration and ATP production rate

We measured the real time mitochondrial oxygen consumption rate (OCR) and extracellular acidification rate (ECAR) using the Seahorse XFe96 Extracellular Flux Analyzer following the manufacturer’s instructions of the Seahorse XF Cell Mito Stress Assay Kit (Cat. No. 103015-100). Briefly, SK-N-SH cells were cultured in a T75 flask in low glucose Dulbecco’s modified Eagle medium (DMEM) (Cat. No. 11-885-092, Thermo Fisher, 10% fetal bovine serum and 1% PS). Cells were seeded into an XF96 Cell Culture Microplate (Cat. No. 102601-100, Agilent Technologies) at 20K cells per well. After 24 h, the cultured media was replaced with serum-free DMEM media. Cells were pretreated (30 min) with mitoTEMPO (2 μM), a mitochondria-targeted antioxidant [mitoTEMPO, 2-(2,2,6,6-tetramethylpiperidin-1-oxyl-4-ylamino)-2-oxoethyl)triphenylphosphonium chloride; Cat. No. SML0737, Sigma-Aldrich] and then incubated with AGEs (100 μg/ml) in the presence of mitoTEMPO for 24 h, or directly treated with AGEs in the absence of mitoTEMPO for 24 h, or treated with BSA (100 μg/ml) as a control. Prior to the assay, cells were washed with assay media (Seahorse DMEM, pH 7.4, containing 10 mM glucose and 2 mM glutamine), and they were incubated in assay media containing BSA, AGEs or AGEs + mitoTEMPO at 37°C (non-CO₂ incubator) for 1 h. MitoTEMPO was added 15 min prior to the stress. The OCR and ECAR were measured in response to oligomycin (2.5 μM), FCCP (2.0 μM) and rotenone/antimycin A (0.5 μM) treatment. The rate of mitochondrial and glycolytic ATP production was determined using the Seahorse XFe96 Extracellular Flux Analyzer, following the manufacturer’s instructions of the Seahorse XF Real-time ATP Rate Assay Kit (Cat. No. 103592-100). Briefly, after SK-N-SH cells were cultured and treated under the same conditions as above, the rate of mitochondrial and glycolytic ATP production was measured upon the sequential addition of oligomycin (2.5 μM) and rotenone/antimycin A (0.5 μM) to the experimental cells. The raw data were analysed using Seahorse Analyzer and GraphPad Prism 9 software.

Mitochondrial reactive oxygen species production and membrane potential in brain slices

Mice were sacrificed by cervical dislocation followed by decapitation, and the entire brain was extracted subsequently. Transverse brain (cerebral cortex and hippocampus) slices (∼300 μm) were obtained by using a tissue-chopper (Electron Microscopy Sciences), maintained in an interface chamber (Harvard Apparatus) at 29°C and perfused with artificial CSF (ACSF) continuously bubbled with 95% O₂ and 5% CO₂. The ACSF contained 124.0 mM NaCl, 4.4 mM KCl, 1.0 mM Na₂HPO₄, 25.0 mM NaHCO₃, 2.0 mM CaCl₂, 2.0 mM MgCl₂ and 10.0 mM glucose. Coronal brain slices were pretreated with mitoTEMPO (2 μM) for 30 min and then incubated with AGEs (100 μg/ml) for 1 h in the presence of mitoTEMPO, or directly treated with AGEs without mitoTEMPO for 1 h, to determine the effect of mitoTEMPO on AGEs-induced mitochondrial reactive oxygen species (ROS). After treatment, brain slices were stained with MitoSOX Red (2.5 μM) or tetramethylrhodamine methyl ester TMRM (100 nM) and DRAQ5 (1:100) for nuclear staining (blue) in living brain slices. MitoSOX Red (Molecular Probes) is a unique fluoroprobe with high selectivity for the detection of superoxide production in the mitochondria of living cells.³¹ The stained brain slices were washed three times with PBS, followed by imaging under a confocal microscope with a 20× objective (AXR Confocal Microscope System, Nikon). Fluorescent signals were quantified using MetaMorph software.

ROS in brain slices were also measured by electron paramagnetic resonance (EPR) as described.³²

Preparation of primary cultured hippocampal neurons

Mouse pups (postnatal Days 0–1) were decapitated. Hippocampi removed by micro-dissection were trypsin-treated (0.25%) and triturated before plating in high glucose DMEM (10% fetal calf serum) (Invitrogen) onto glass cover slips coated with poly-D-lysine (20 μg/ml) (Sigma-Aldrich) as described.^33,34 Cells were maintained in an incubator at 37°C with 5% CO₂. The next day, the medium was replaced with Neurobasal A medium supplemented with 2% B27 and 2 mM GlutaMAX (Life Technologies).³⁵ Cells were then stored in the incubator for 10 days, followed by replacement of half of the medium with fresh B27-Neurobasal A medium containing 50 or 100 μg/ml AGEs, or vehicle, and kept for 24 h. Primary hippocampal neurons [days in vitro (DIV) 12–13] were used for patch clamp experiments.

Electrophysiologic studies on hippocampal slides from transgenic mice

For further details refer to the Supplementary material.

Whole-cell patch clamp studies

Patch clamp studies in neuronal cultures were performed as previously described.^36,37 Briefly, for patch-clamp recording in cell cultures, cover slips were transported from the incubator to the recording chamber in a cell culture dish containing recording bath solution comprised of (in mM): 119 NaCl, 5 KCl, 20 HEPES, 30 glucose, 2 CaCl₂ and 2 MgCl₂. The osmolarity was adjusted to ∼330 mOsm with sucrose and the pH adjusted to 7.3 with 10 N NaOH. We acquired whole-cell patch clamp recordings by using a MultiClamp 700B amplifier and Clampex data acquisition software 14.2 (Molecular Devices) at room temperature. A glass pipette of 5–8 mΩ was fabricated and filled with electrode solution (in mM): 130 potassium-gluconate, 10 KCl, 5 HEPES, 5 MgCl₂, 0.06 CaCl₂, 0.6 EGTA, 2 MgATP, 0.2 Na₂GTP and 20 phosphocreatine. The osmolarity was adjusted to 310 mOsm and the pH adjusted to 7.1 with KOH. Miniature excitatory postsynaptic currents (mEPSCs) were recorded in the presence of 100 μM picrotoxin and 0.5 μM TTX. The threshold for mEPSC event detection was set at 5 pA. Signals were filtered at 2 kHz, stored and analysed offline using Clampfit 10 (Molecular Devices).

Morris water maze

Mice were subjected to the Morris water maze (MWM) hidden platform test as described in our previous studies.¹⁸ Briefly, mice were trained for five consecutive days with four trials per mouse per day. On the last day, a probe trial was performed to assess the spatial memory of the mice. Traces of mice were recorded and data were analysed using HVS Image 2014. All animals were individually coded, and investigators were blinded to the mouse genotypes for the duration of behavioural testing.

Statistical analysis

All data were expressed as the mean ± standard error of the mean. Statistical significance was determined by unpaired t-test or ANOVA test (one-way or two-way) according to the number of independent variables or to the levels of two categorical variables using GraphPad Prism 9 or StatView software (version 5.0.1, Berkeley, CA). Fisher's multiple comparisons test was performed for post hoc comparisons in StatView. P < 0.05 was considered to be significant. Pearson correlation analysis was performed using GraphPad Prism 9 software to measure the strength and direction of the linear relationship between two variables.

Results

Precursor of advanced glycation end product accumulation in synaptic mitochondria with brain ageing

To determine whether ageing altered the accumulation of glycated adducts in the synaptic and non-synaptic mitochondria in mouse brains, synaptic and non-synaptic mitochondria were prepared from mouse cortices as previously described.^26,38 The purity of the synaptic mitochondria was ascertained by the enrichment of mitochondrial markers (TOM20 and VDAC) and the relative absence of endoplasmic reticulum (calnexin) and lysosomal (LAMP-1) and neuronal (synaptophysin) markers (Fig. 1A). We then measured levels of MG, one of the best-known and most potent precursors of AGEs, in synaptic and non-synaptic mitochondria. To examine the age-dependent accumulation of mitochondrial MG and its association with AGE accumulation and mitochondrial dysfunction, mice were analysed at the ages of 3, 6, 12, 20 and 30 months. Synaptic and non-synaptic mitochondrial fractions were subjected to HPLC to detect MG. The MG concentrations (nM) of synaptic mitochondria at 6 and 12 months were 6.2 ± 0.11 and 14.6 ± 0.50, respectively, whereas the MG concentrations of non-synaptic mitochondria were less than 1 nM (Fig. 1B; P < 0.05). By 20 to 30 months of age, MG in synaptic mitochondria reached 16.0 ± 0.80 and 26.2 ± 1.2, respectively. Clearly, as early as 6 months of age, MG levels were significantly higher in synaptic mitochondria than in non-synaptic mitochondria (Fig. 1B–D). This was further confirmed by the significant positive correlation of MG levels with increasing age (Fig. 1C). MG content in the mitochondrial fraction was based on MG standard by HPLC (Supplementary Fig. 1). These results indicated that AGE-related metabolites accumulate progressively to a great degree in synaptic mitochondria in an age-dependent manner.

Figure 1.

Accumulation of methylglyoxal in the synaptic mitochondria of ageing mice. (A) Immunoblots of synaptosome and synaptic mitochondrial fractions from mouse brain were carried out to detect neuronal presynaptic marker (synaptophysin), mitochondria (VDAC and Tom20), lysosome (Lamp-1) and endoplasmic reticulum (ER; calnexin). (B) Determination of methylglyoxal (MG) levels in the synaptic and non-synaptic mitochondria of mice using high-performance liquid chromatography (HPLC). (C) Correlation between MG levels and mice ages. (D) Representative HPLC chromatograms that identify MG. n = 3–8 mice per group. *P < 0.05. ^#Significant difference (P < 0.05) among synaptic and non-synaptic mitochondria groups at the same age.

Advanced glycation end product accumulation in synaptic mitochondria with brain ageing

We next evaluated whether synaptic mitochondrial accumulation of MG correlated to MG-derived AGEs, CML and total AGEs with ageing. To this end, protein extracts of synaptic and non-synaptic mitochondrial fractions from mouse brains of different ages were analysed using ELISA to measure total AGEs, CML and MG-AGEs, respectively. Interestingly, parallel to MG accumulation, levels of MG-AGEs increased significantly in both synaptic and non-synaptic mitochondria with advancing ageing (Fig. 2A). However, at 20 and 30 months of age, MG-AGE levels in synaptic mitochondria (35.3 ± 3.4 and 54.0 ± 3.3 ng/mg protein) were significantly higher than those in non-synaptic mitochondria (19.9 ± 1.8 and 30.3 ± 2.2). These results indicated that MG-dependent accumulation of AGEs was predominant in synaptic mitochondria. Similarly, at 20 and 30 months of age, CML concentration levels (ng/mg protein) were higher in synaptic mitochondria (25.2 ± 2.3 and 48.5 ± 1.9) than in non-synaptic mitochondria (12.7 ± 2.5 and 23.3 ± 1.5) (Fig. 2B). As a result, total AGE accumulation in both non-synaptic and synaptic mitochondria increased with age, but AGE levels in synaptic mitochondria were further elevated at 30 months of age (Fig. 2C). Significant positive correlation of AGEs, CML and MG-AGE levels with increasing age further supported the susceptibility of synaptic mitochondria to the accumulation of MG and AGEs (Fig. 2D–F). These results indicated that AGE-related metabolites accumulate progressively and to a high degree in synaptic mitochondria in an age-dependent manner. To further confirm the presence of AGEs in synaptic mitochondria, an immunogold electron microscope was used to examine the intact mouse brains using a specific primary anti-AGE antibody followed by a secondary antibody conjugated to gold particles as described.³⁹ AGE immunogold particles were observed to localize to presynaptic and postsynaptic mitochondria in mouse brain (Fig. 2G and H). When the AGE antibody was omitted, the specific AGE staining pattern disappeared (Fig. 2I).

Figure 2.

Accumulation of advanced glycation end-products in the synaptic mitochondria of ageing mice. (A–C) Quantification, using ELISA, of methylglyoxal-advanced glycation end-products (MG-AGEs) (A), carboxymethyl-lysine (CML) (B) and total AGEs (C) in synaptic and non-synaptic mitochondria from mouse brain. n = 3–8 mice per group. *P < 0.05, **P < 0.01, ***P < 0.001. ^#Significant difference (P < 0.05) among synaptic and non-synaptic mitochondria groups at the same age. (D–F) Correlation between synaptic and non-synaptic mitochondria AGE levels and mouse age. (G–I) Electron microscopy shows post-embedding immunogold labelling of synapses and mitochondria in cortex using an anti-AGE antibody. AGE-immunogold particles (12 nm) aggregated in the synaptic mitochondria (G) and the postsynaptic dendrites (H) of 21-month-old mice. A control without AGE antibody (I). Asterisks mark the active zones of synapses. Scale bar = 250 nm.

Accumulation of advanced glycation end products results in synaptic mitochondrial dysfunction

To determine if synaptic mitochondrial accumulation of AGEs correlates with abnormal mitochondrial function, we first investigated mitochondrial respiratory function and energy metabolism by measuring the activities of mitochondrial respiratory chain complexes I (COX I) and IV (COX IV) and ATP levels in ageing mice. COX I or COX IV activity was significantly reduced in synaptic and non-synaptic mitochondrial fractions in 20- and 30-month-old mice compared with 3-month-old mice. However, the decrease in COX I or COX IV activity was greater in synaptic mitochondria from 20- and 30-month-old mice compared with non-synaptic mitochondria (Fig. 3A and B).

Figure 3.

Age-dependent synaptic mitochondrial dysfunction in mice. (A and B) The activities of complex I (COX I; A) and complex IV (COX IV; B) were determined in isolated synaptic and non-synaptic mitochondria of mice (n = 5 per group). *P < 0.05, **P < 0.01, ***P < 0.001. ^#Significant difference (P < 0.05) among synaptic and non-synaptic mitochondria groups at the same age. Representative TMRM staining images (C) and quantification of TMRM staining intensity (D) in brain slices treated with advanced glycation end-products (AGEs; 100 μg/ml) or vehicle for 1 h. n = 7 per group and ***P < 0.001. Scale bar = 25 μm. (E–I) ATP (E) was determined in isolated synaptic and non-synaptic mitochondria of mice (n = 5 per group). *P < 0.05, **P < 0.01, ***P < 0.001. ^#Significant difference (P < 0.05) among synaptic and non-synaptic mitochondria groups at the same age. Correlation of methylglyoxal (MG) or AGEs with ATP levels in synaptic (F and G) and non-synaptic mitochondria (H and I). (J–N) H₂O₂ levels (J) were determined in isolated synaptic and non-synaptic mitochondria of mice (n = 5 per group). *P < 0.05, **P < 0.01, ***P < 0.001. ^#Significant difference (P < 0.05) among synaptic and non-synaptic mitochondria groups at the same age. Correlation of MG or AGEs with H₂O₂ levels for synaptic (K and L) and non-synaptic mitochondria (M and N). Representative MitoSox Red staining images (O) and quantification (P) in AGE-, vehicle- or AGE + mitoTEMPO (2 μM)-treated samples (n = 6 per group). ***P < 0.001. Scale bar = 25 μm. (Q–T) Representative oxygen consumption rate (OCR) graph after injection of oligomycin, FCCP or Rotenone/antimycin (Q). Maximal OCR (R), mitochondria- (S) or glycolysis-derived ATP (T) production rate in human neuronal cells exposed to AGEs in the presence or absence of mitoTEMPO (2 μM). Data are presented as mean ± standard error of the mean (n = 7/group). *P < 0.05, **P < 0.01.

To evaluate the effects of AGEs on hippocampal mitochondria in brain, brain slices were perfused with AGEs (100 μg/ml) in ACSF buffer to maintain live mitochondria as an ex vivo model for real-time in situ assessment of mitochondrial membrane potential. To assess the mitochondrial membrane potential of hippocampal neurons, TMRM (100 nM) dye, a specific marker of mitochondrial membrane potential, was used. TMRM fluorescence was significantly reduced (∼50%) in AGE-treated slices compared with vehicle-treated slices (Fig. 3C and D). In parallel, ATP levels in synaptic mitochondria were significantly lower than non-synaptic mitochondria from 12-, 20- and 30-month-old-mice, respectively, although ATP levels were comparable in 3-month synaptic, non-synaptic and 12-month non-synaptic mitochondria (Fig. 3E). These data demonstrated that mitochondrial respiration and energy metabolism were more deteriorated in synaptic mitochondria than non-synaptic mitochondria. Next, ATP levels were negatively correlated with the accumulation of MG and AGEs using Pearson's correlation (Fig. 3F–I). Clearly, synaptic mitochondria displayed a greater degree of negative correlation between MG/AGE levels and ATP (Fig. 3F and G) as compared to non-synaptic mitochondria (Fig. 3H and I).

Given that mitochondria are the primary sites of ROS generation, and that oxidative stress is strongly associated with impaired mitochondrial respiration, we measured hydrogen peroxidase (H₂O₂) levels in mitochondrial fractions of the brain. H₂O₂ levels were significantly higher in synaptic mitochondria than non-synaptic mitochondria in 12–30-month-old mice, although both synaptic and non-synaptic mitochondria exhibited a significant increase in H₂O₂ levels with age (Fig. 3J). We also used Pearson’s correlation to determine the direction and strength of the linear relationship between MG/AGE accumulation and oxidative stress (H₂O₂) in synaptic and non-synaptic mitochondria (Fig. 3K–N). MG and AGEs showed a significant and positive correlation with H₂O₂ in both synaptic (Fig. 3K and L) and non-synaptic mitochondria (Fig. 3M and N), whereas the synaptic mitochondrial pool of MG and AGEs was to a great degree associated with H₂O₂ levels. These data suggested a link between mitochondrial MG/AGE accumulation and oxidative stress. Synaptic mitochondria are prone to the accumulation of oxidative stress during ageing.

Next, we evaluated the AGE-dependent elevated oxidative stress in the cortical and hippocampal regions in living brain slices as an ex vivo model of mice. Brain slices were perfused with AGEs in ACSF buffer for real-time in situ assessment of oxidative stress, and then incubated with MitoSOX (2.5 μM). Images were captured under a confocal microscope. MitoSOX staining signals were significantly elevated in the hippocampus and cortex of AGE-treated slices as compared to vehicle-treated slices (Fig. 3O and P). Scavenging mitochondrial ROS by the application of the mitochondria-targeted antioxidant mitoTEMPO abolished the MitoSOX staining signal, suggesting that AGEs mediate excess mitochondrial oxidative stress. Overall, the in vivo and ex vivo results are in a very good agreement that the mitochondrial AGEs are closely related to aberrant mitochondrial function and oxidative stress, to which synaptic mitochondria are more susceptible.

Finally, we assessed the real-time mitochondrial function using a Seahorse assay to confirm the effects of AGEs and related oxidative stress on neuronal mitochondria. The real time OCR was determined upon sequential exposure to oligomycin (2.5 μM), FCCP (2.0 μM) and rotenone/antimycin (0.5 μM) using Seahorse XF96e Analyzer (Fig. 3Q). Treatment with AGEs impaired mitochondrial respiration in live human neuronal cells as shown by the reduction in maximal OCR (Fig. 3Q and R). We also observed a significant reduction in mitochondrial and glycolytic ATP production in the presence of AGEs, as determined by the Seahorse XF96e analyzer in response to oligomycin (2.5 μM) and rotenone (0.5 μM) treatment (Fig. 3S and T). Interestingly, these detrimental effects were almost completely reversed by mitoTEMPO treatment (Fig. 3R–T). Thus, suppression of mitochondrial toxic glycated products and associated oxidative stress could protect against mitochondria injury relevant to brain ageing.

Neuronal GLO1 eliminates AGE accumulation and restores function in synaptic mitochondria

Given that GLO1 activity was significantly reduced with ageing (Supplementary Fig. 2), we next determined whether clearance of AGEs by restoring/enhancing GLO1 expression/activity in cortical neurons would reduce the synaptic (neuronal) mitochondrial pool of AGEs. To this end, we generated transgenic mice overexpressing neuronal GLO1 (Tg Glo1) under the thy-1 promoter (Supplementary Fig. 3A and B). GLO1 mice expressed significantly higher levels of GLO1 in cortical neurons than non-transgenic (non-Tg) mice (Supplementary Fig. 3C–F). Accordingly, GLO1 activity was elevated in Tg Glo1 mice (Supplementary Fig. 3F). This newly created neuronal GLO1 mouse is an appropriate mouse model to investigate the effects of AGE metabolism/clearance via GLO1 on endogenous and exogenous AGE accumulation in vivo, as well as mitochondrial and cognitive function. Indeed, we observed a significant reduction in naturally produced endogenous MG and toxic species of AGEs (MG-AGEs and CML) in the synaptic mitochondria of GLO1 mice compared with non-Tg mice (Fig. 4A–C). Accordingly, the activity of mitochondrial respiratory enzyme (COX 1; Fig. 4D), COX IV and ATP levels were significantly elevated in GLO1 mice (SupplementaryFig. 3G and H).

Figure 4.

Neuronal GLO1 overexpression reduces the accumulation of synaptic mitochondrial methylglyoxal and advanced glycation end products (AGEs) in aged mice and mice fed with an AGE-rich diet. Non-transgenic (non-Tg) and GLO1 mice were fed with an AGE-free (AGE−) or AGE-rich (AGE+) diet for 12 months starting at 9 months of age. (A) Methylglyoxal (MG) levels in the synaptic mitochondria of non-Tg and GLO1 mice fed with an AGE− or AGE+ diet were determined using high-performance liquid chromatography. (B and C) Quantification of immunodot blots of MG-AGEs (B) and carboxymethyl-lysine (CML) (C) relative to Hsp60 from the synaptic mitochondrial fractions of AGE− and AGE+ diet-fed mice. Bottom: Representative immunoblots of synaptic mitochondrial fractions, indicating MG-AGEs, CML and Hps60. Hsp60 was used as the mitochondrial protein loading control. n = 3 mice. Data are presented as mean ± standard error of the mean. (D–F) Complex I (COX I) activity (D), representative electron paramagnetic resonance (EPR) spectra (E) and reactive oxygen species (ROS) levels (F) in mice fed with AGE− or AGE+ diets. n = 3–5 mice per group. (G) Representative image of MitoSOX staining in non-Tg and GLO1 hippocampaal slices treated with vehicle or AGEs (100 μg/ml). Scale bar = 50 μm. Bottom row: Enlarged images of areas outlined by the white dotted lines. (H) Quantification of MitoSOX fluorescence signals in non-Tg and GLO1 hippocampal slices treated with vehicle or AGEs (100 μg/ml). n = 6 from three mice per group. (I–K) Effect of AGEs on ROS level and GLO1 activity. Representative EPR spectra (I) and ROS levels (J) in non-Tg and GLO1 brain slices treated with vehicle or AGEs (100 μg/ml). Data are presented as mean ± standard error of the mean. n = 8–13 sections from 3–5 mice per group. Both male and female mice were used in the experiment. (K) GLO1 activity was reduced in AGE-treated (100 μg/ml) human neuroblastoma cells (SK-N-SH). Treatment with mTEMPO (2 μM) completely restored GLO1 activity. Data were analysed by one-way ANOVA followed by Fisher’s post hoc test. *P < 0.05, ns = no significance. n = 3 per group.

In view of the significance of the exogenous AGEs on neuronal function and the pathological role of AGE-rich diet, we next determined the effects of neuronal GLO1 on AGE diet-mediated mitochondrial and cognitive function. To this end, mice were fed with an AGE diet for 12 months starting from 9 months of age. Consistent with our previous report,¹⁸ AGE diet-fed mice exhibited a significant increase in the accumulation of AGEs and mitochondrial defects. Importantly, the AGE precursor MG, and the levels of AGEs, including MG-derived AGE and CML, were robustly reduced by 60%–80% in synaptic mitochondria in AGE diet-fed GLO1 mice compared to AGE diet-fed non-Tg mice (Fig. 4A–C). Evidently, GLO1 overexpression protected against AGE diet-induced mitochondrial dysfunction and oxidative stress, as shown by fully restored COX I activity (Fig. 4D) and suppression of cerebral and mitochondrial ROS using quantitative specific and highly sensitive EPR (Fig. 4E and F). Furthermore, aged GLO1 mice not only had 74% reduced accumulation of endogenous AGEs but also decreased ROS and increased COX I activity compared with non-Tg mice (Fig. 4A–F). The in vivo results were further supported by an ex vivo model in which brain slices from non-Tg mice perfused with AGEs displayed abundant ROS (MitoSOX signals) in cortical (Supplementary Fig. 4A and B) and hippocampal regions (Fig. 4G and H). Notably, the hippocampus is more vulnerable to AGE-induced oxidative stress than the cortex, which may be directly related to LTP impairments in hippocampal neurons. Importantly, the increased GLO1 expression/activity attenuated AGE-mediated oxidative stress in cortical and hippocampal regions (Fig. 4G–K). Scavenging mitochondrial target ROS by application of mitoTEMPO fully rescued GLO1 activity in AGE-treated neuronal cells (Fig. 4K), suggesting the link between AGE-involved oxidative stress and GLO1 function. Accordingly, GLO1 mice were resistant to AGE-induced ROS in an ex vivo brain slice perfusion with AGEs by EPR measurement (Fig. 4I and J). These results indicated a strong correlation of AGEs/ROS with GLO1 function. Neuronal GLO1 sufficiently eliminates endogenous and exogenous AGE accumulation and improves mitochondrial function.

Neuronal GLO1 rescues AGE-induced synaptic function and improves learning and memory

Because of the relevance of synaptic mitochondrial function to synaptic activity, we next assessed whether GLO1-improved mitochondrial function was reflected in synaptic function by measuring synaptic plasticity and pre- and post-synaptic activity. We recorded LTP in CA1 neurons of hippocampal slices in the presence and absence of AGE treatment. AGEs significantly impaired LTP, whereas slices from GLO1 mice showed complete recovery from AGE-induced LTP deficits (Fig. 5A). No differences in the input and output of basal neurotransmission were found between non-Tg and GLO1 mice (Supplementary Fig. 5), suggesting no effect of GLO1 overexpression on the physiological synaptic baseline.

Figure 5.

Neuronal GLO1 rescues advanced glycation end product-induced deficits in synaptic long-term potentiation and spontaneous neurotransmitter release, as well as learning and memory. (A) Advanced glycation end products (AGEs; 150 μg/ml) were perfused for 20 min prior to theta-burst stimulation. n = 10–15 slices per group. ***P < 0.01. (B–F) Effect of GLO1 on miniature excitatory postsynaptic currents (mEPSCs). (B and E) Treatment with AGEs (100 μg/ml) significantly reduced the frequency of mEPSCs in non-transgenic (non-Tg) but not GLO1 mouse neurons. (D). Representative traces of mEPSCs recorded from nonTg or GLO1 hippocampal neurons pre-treated with 50 μg/ml or 100 μg/ml AGEs for 24 h. Under all conditions, the amplitudes of the mEPSCs did not change (C and F). Overall, 30–36 neurons from at least three independent litters of animals were recorded in each group. Data are shown either as scatter plots with mean ± standard error of the mean (*P < 0.05, **P < 0.01, ***P < 0.001, ns = no significance) or cumulative probability plots (E and F). (G–J) Effect of GLO1 on AGE+ diet-impaired learning and memory. AGE− and AGE+ diet-treated GLO1 mice overexpressing neuronal GLO1 and non-Tg mice at 22 months of age were subjected to the Morris water maze (MWM) test. (G) Escape latency during the MWM hidden platform task. (H) Time spent in the quadrant with the hidden platform during the probe test. (I) Mean number of crossings of the target during the probe test. n = 7–10 mice per group. *P < 0.05, **P < 0.01, ***P < 0.001. (J) Representative tracing images from non-Tg or GLO1 mice fed with AGE− or AGE+ diet.

To determine the effect of GLO1 on AGE-induced changes in pre- and postsynaptic activity, we measured spontaneously occurring mEPSCs from primary cultured hippocampal neurons derived from GLO1 and non-Tg mice through whole-cell patch-clamp recording. Treatment of non-Tg hippocampal neurons with AGEs (100 μg/ml) displayed a significant decrease in the frequency of mEPSCs compared with those treated with control BSA (Fig. 5B and D). The distribution of mEPSC inter-event intervals was shifted towards longer intervals in AGE-treated neurons (Fig. 5E), confirming an overall tendency to lower mEPSC frequency. Importantly, AGE-reduced mEPSC frequency was fully restored in GLO1-expressing neurons (Fig. 5B, D and E). These changes in mEPSC frequency were not accompanied by changes in mEPSC amplitude between non-Tg and GLO1 neurons with or without AGE treatment (Fig. 5C and F). Altogether, these changes suggested an alteration in the probability of spontaneous neurotransmitter release from pre-synaptic but not post-synaptic sites (mEPSC traces are shown in Fig. 5D). Hence, AGE treatment impairs spontaneous synaptic transmission, possibly via a reduction the number of active synapses, but does not alter the postsynaptic AMPA receptor activity of each single synapse. These results suggested that increasing GLO1 expression is effectively beneficial in reversing AGE-mediated toxicity and restoring memory and cognitive function.

The synapse plays a crucial role in the consolidation of memory, and synaptic degeneration results in diminished cognitive ability. Thus, we next assessed the effect of GLO1 on AGE-induced learning and memory defects. Mice were subjected to the Morris water maze, a test of long-term memory. Compared to normal diet-fed mice, AGE-fed mice exhibited significantly longer escape times during the 5-day training session (Fig. 5G), reduced numbers of crossings of the target area and less time spent in the target quadrant during the probe trial (Fig. 5H–J). Conversely, AGE-fed GLO1 mice displayed substantially shorter escape times and increased numbers of crossings and time spent on the platform relative to AGE diet-fed non-Tg mice (Fig. 5G–J). Interestingly, GLO1 mice also exhibited a significant increase in time spent on the platform compared to non-Tg mice fed with an AGE-diet. These data indicated that the enhancement of neuronal GLO1 not only improves mitochondrial and synaptic function but also protects against the detrimental effect of an AGE-enriched diet on learning and memory.

Discussion

Mitochondrial dysfunction and synaptic injury are early pathological features of neurodegeneration, AD and age-related memory loss.^1,40-44 Synaptic mitochondrial dysfunction occurs during ageing, correlates with cognitive decline and is an early target of pathogenic amyloid-β and tau in AD.^26-29,40,45 The causes and consequences of early synaptic mitochondrial perturbation remain to be determined. In the present study, we demonstrated for the first time that synaptic mitochondria are reservoirs for the accumulation of AGE-related toxic metabolites, including a highly reactive dicarbonyl species MG in ageing and AGE-enriched brain. High AGEs in synaptic mitochondria lead to mitochondrial stress and dysfunction, resulting in cognitive decline. Thus, synaptic mitochondrial glycation could initiate/accelerate mitochondrial dysfunction and synaptopathology. Clearance or detoxification of these toxic metabolites is essential to maintain the integrity of mitochondria and synaptic function.

First, we have clearly demonstrated that AGEs and related toxic metabolites accumulate to a greater extent in synaptic mitochondria than in non-synaptic mitochondria in an age-dependent manner. Notably, MG, one of the most important AGE precursors, significantly accumulated in synaptic mitochondria as early as 6 months of age and progressively increased with age. Accordingly, synaptic mitochondria enriched for MG/AGEs displayed detectable pathological and functional alterations, including decreased activity of key mitochondrial respiratory enzymes, reduced ATP levels, and elevated oxidative stress. These deleterious effects on synaptic mitochondria preceded or worsened the damage to non-synaptic mitochondria. Additionally, mitochondrial membrane potential and mitochondria-derived ROS were significantly decreased and increased, respectively, in the presence of AGEs. These findings suggest that synaptic mitochondria are more susceptible to AGE-induced injury than non-synaptic organelles and that synaptic mitochondrial stress is an early pathological event in age-related dementia.

Although AGEs accumulation is part of the normal ageing process, when significantly accelerated, such as in diabetes, atherosclerosis, cardiovascular diseases and neurodegenerative diseases, including AD, increasing amounts of AGEs can initiate/accelerate the development and/or disease progression. Increased AGE levels are associated with cognitive decline in older adults with and without diabetes.¹¹ In AD, AGEs induce crosslinking of amyloid-β and hyperphosphorylated tau, which are hallmarks of AD pathology.^15,16,46,47 Application of AGEs to brain slices perturbs mitochondrial function and energy metabolism and increases ROS generation¹. Additionally, excessive AGEs accumulation disrupts synaptic transmission and reduces long-term potentiation,¹⁹ leading to cognitive decline during ageing and accelerated progression from mild cognitive impairment to AD.^1,11 MG derivatives are also elevated in normal individuals consuming AGE-rich diets and in the brains of mice fed with an AGE-rich diet,^18,48 resulting in decreased mitochondrial respiratory enzyme activity and ATP levels and increased oxidative stress. AGEs in food are thereby causative factors for development and/or progression of brain pathologies. Thus, the accumulation of AGEs in synapses and synaptic mitochondria could serve as danger signals that trigger and/or promote mitochondrial and synaptic degeneration.

The formation of AGEs occurs both endogenously and exogenously. Age-related dementia, metabolic changes and cognitive decline may be causally linked to foods with high levels of AGEs.^12,18,49,50 Accumulating evidence points to the importance of foods as an exogenous source of AGEs.^12,42,51 Human exposure to AGEs is widespread, especially because most Western foods are processed at high temperatures and are therefore rich in AGEs. For example, through barbecuing, grilling, roasting, frying, and toasting, AGE content may be 10–100 times higher than that of uncooked foods, resulting in potentially harmful AGE intake.^52,53 Consequently, modern dietary changes include excessive nutrient-bound AGEs, such as neurotoxic MG derivatives. It is essential to determine whether an increased dietary intake of AGEs accelerates and exacerbates mitochondrial function, impairs learning and memory, and whether elimination of AGEs alleviates these detrimental effects. Mice fed with an AGE-rich diet provide an appropriate model to further determine the direct effect of GLO1 on AGE-mediated mitochondrial and cognitive dysfunction and AGE metabolism. The results generated from mice fed an AGE-rich diet address the key unexplored questions outlined earlier. In terms of biological age, 18–24 month-old mice are equivalent in age to 56–69 year-old humans. Age is the biggest known risk for AD, and most people with late-onset AD (sporadic AD) develop the disease and cognitive decline when they are 65 or older, and the risk of developing AD further increases with ageing over 65. Additionally, age-related metabolic changes significantly promote the accumulation of AGEs. Specifically, impaired GLO1-mediated clearance of MG/AGEs, as indicated by declining GLO1 function, is noted with advancing age, with significantly lower GLO1 activity levels in mice at 20 months of age compared to younger mice. Thus, it is rational to use 21-month-old mice to investigate the effect of endogenous and exogenous AGEs on mitochondrial and cognitive function. Using mice at this age addresses both whether enhancing GLO1 function affords long-term protection against chronic exposure to an AGE-rich diet and eliminates endogenous build-up of toxic metabolites in ageing mice. We demonstrated that mice fed a high AGE diet had significantly higher levels of MG/AGEs in synaptic mitochondria compared to mice fed a normal diet. In contrast, synaptic mitochondrial pools of MG and toxic AGE species were robustly reduced in GLO1 mice fed an AGE diet compared to the MG/AGEs levels observed in nonTg mice fed with an AGE diet. Accordingly, AGEs/GLO1 mice do not display mitochondrial respiratory failure and oxidative stress. These results indicate the impact of GLO1 in the clearance/elimination of MG/AGEs and the maintenance of mitochondrial function.

GLO1 forms an important innate defence system as the first key, limiting enzyme for detoxifying MG and preventing the accumulation of AGEs during AGE metabolism in the glycolytic pathway. Notably, scavenging mitochondrial ROS by applying mitochondria-targeted antioxidant fully restored the GLO1 activity that had been reduced by AGEs, suggesting a link between oxidative stress and GLO1 dysfunction. To address whether clearance and detoxification of MG/AGEs through gain-of-GLO1 function would attenuate the deleterious effects of AGEs on mitochondrial, synaptic, and cognitive function, we generated novel transgenic mice overexpressing GLO1 in cortical neurons. This novel animal model displayed a significant increase in GLO1 expression compared to age-matched non-Tg mice. The activity of GLO1 was increased in the brain and mitochondria. As a result, GLO1 mice not only showed reduced levels of endogenous (naturally produced) MG/AGEs in synaptic mitochondria, but also enhanced mitochondrial respiration through increased activity of mitochondrial COX I and COX IV, elevated ATP production and reduced ROS levels. Thus, this well-characterized GLO1 mouse strain provides a unique mouse model for determining the role of age-related metabolic changes in brain ageing and an AGE-rich milieu relative to mitochondrial and cognitive function. To maintain synaptic transmission, synapses are sites of high energy demand and extensive calcium fluctuations.^26,54 Given that AGEs cause significant mitochondrial defects and that synaptic integrity is critical for brain function and memory, we propose that AGE-mediated defects in synaptic plasticity result from insufficient mitochondrial integrity. Indeed, our present in vivo study demonstrated that when chronically consuming a diet rich in AGEs, GLO1 mice are able to protect against AGE-induced cognitive decline, as observed in behaviour tests, which demonstrated shortened platform-seeking latency, a reduced number of platform crossings and increased time spent in the target quadrant. Additionally, neuronal expression of GLO1 reduced endogenous MG/AGEs accumulation and enhanced cognition, as revealed by the increased time spent in the target area of the MWM. Furthermore, GLO1 expression conferred resistance to AGE-mediated decline in LTP and presynaptic activity in hippocampal neurons, as shown by an increased mEPSC frequency. These results further support the protective role of enhanced GLO1 function against AGE-mediated synaptic failure and memory loss.

In summary, we provide substantial evidence that synaptic mitochondria exhibit age-dependent progressive accumulation of AGE-related metabolites and mitochondrial alterations to a greater extent than non-synaptic mitochondria. Such accumulation is positively correlated with mitochondrial perturbation and oxidative stress. Using novel neuronal GLO1 mice, we explored the in vivo role of GLO1 in AGE metabolism and the crosstalk between age-related metabolic changes and mitochondrial and synaptic stress. Gain-of-function of GLO1 significantly eliminates synaptic mitochondria glycation, thereby rescuing mitochondrial function, suppressing ROS production, and improving learning and memory in AGE-challenged mice. Accordingly, increased expression of neuronal GLO1 protects against AGE-induced deficits in synaptic transmission and plasticity. We propose that synaptic mitochondria are an early and primary target of AGE toxicity and that synaptic mitochondrial AGEs could serve as endogenous danger signals to initiate/promote mitochondrial perturbation and oxidative stress, leading to synaptic failure and cognitive decline. Thus, augmenting GLO1 function to scavenge and detoxify toxic metabolites could be a potential therapeutic strategy against age-related mitochondrial and neuronal degeneration and cognitive decline.

Data availability

The datasets during and/or analysed during the current study available from the corresponding author on reasonable request.

Funding

This study was supported by grants from the National Institute on Aging (R56AG064934, RF1AG081575, R01AG083340, RF1AG077848, R01AG061324).

Competing interests

The authors declare that they have no competing interests.

Supplementary material

Supplementary material is available at Brain online.