Age-related changes in synapse ultrastructure and neuroprotective effect of dietary curcumin in the lateral prefrontal cortex layer 3 of the rhesus monkey

Abstract

Elucidating the impact of aging on the structure and function of neurons is key to understanding the mechanisms underlying synaptic dysfunction and ensuing susceptibility to age-related cognitive decline. The role of structural alterations in the lateral prefrontal cortex (LPFC) has been extensively addressed. In addition, numerous studies point to the importance of inflammation and the increase of oxidative stress during aging, with mitochondria as one of the key cellular organelles involved. Here, we used 3D high-resolution serial block-face scanning electron microscopy to visualize the ultrastructure of synapses on pyramidal neurons in area 46 of the LPFC in rhesus monkeys at different stages across their adult lifespan. Our results revealed a general loss of synapses with age, mainly driven by the loss of asymmetric axospinous synapses. We observed a larger bouton volume but not larger spine or postsynaptic density (PSD) surface in the aged group compared to all other groups, along with a weaker correlation between spine and synaptic size. Additionally, we found morphological changes in mitochondria in the aged compared to middle-aged and young monkeys. Altogether, our data show ultrastructural changes that suggest an improper synaptic scaling and possible mitochondrial dysfunction that might take place after middle-age. We studied the impact of curcumin as a long-term dietary supplement and found that it ameliorated some of these age-related changes at middle-age by preserving the spine and PSD morphology and their size relationship, and also mitochondrial morphology, which might allow for maintaining synaptic function during aging, resulting in a delayed cognitive decline.

Introduction

Elucidating the mechanisms that render synapses dysfunctional during normal aging is key to understanding variability in individual age-related impairment in cognitive function that occurs in human and non-human primates (NHP) [1–5]. Structural and functional alterations in layer 3 pyramidal neurons in the lateral prefrontal cortex (LPFC) have been extensively characterized in NHP. Retrograde tract-tracing and dye-filling of cortical pyramidal neurons in macaques and other species have also shown a reduction in dendritic length and segment numbers in both apical and basal dendrites with age [6–8]. Later studies in cognitive aging in human and NHP confirmed that synapse loss, that accompanies the decrease in dendritic spines in LPFC, is associated with cognitive decline [9–14]. Age-related increases in the size of thin spines was correlated with a worse performance on the cognitive delayed non-matching to sample (DNMS) task [15] whereas sizes of mushroom spines do not correlate with behavior and seem to remain more stable during aging [11, 12, 14]. Electron microscopy (EM) studies, that assessed changes in synaptic density and synapse types, confirmed that glutamatergic axospinous synapses exhibit the largest age-related decrease in the LPFC layer 3 [9, 11, 16], compared to overall loss of synapses from other LPFC layers. Trying to explain the differences among cortical layers, electrophysiological experiments from area 46 showed that the basic electrophysiology of layer 3 pyramidal neurons changes significantly with age, and is associated with decreased cognitive performance [17–24].

In addition, studies have focused on the role of age-related increases in oxidative stress and inflammation [25, 26]. Mitochondria are especially vulnerable to oxidative stress damage, directly affecting their ability to generate energy [27–32]. Thus, age-related changes in mitochondrial morphology have been described to take place in NHP during aging [33, 34]. At the synapse, these alterations in mitochondrial structure and length are likely to affect their function and the performance on working memory tasks in NHP [33, 34].

Given the consequences of oxidative stress on cognition, there is a growing interest in the potential of nutraceuticals with antioxidant and anti-inflammatory properties [35–37]. Curcumin is a polyphenol component of turmeric [38] that has been reported to slow or reverse the negative effects of the oxidative stress on the brain during aging in numerous studies using rodent models and human monkeys [39–41]. Our group has also recently shown that dietary curcumin improves working memory performance [42] and ameliorates age-associated neuronal hyperexcitability in vitro, in middle-aged rhesus monkeys compared to control middle-aged and aged monkeys [23].

In the present study, we investigate changes in synapse density and ultrastructure in middle-aged rhesus monkeys related to cognitive decline and neurodegeneration during aging, as well as the effects of dietary curcumin supplementation. We used three-dimensional (3D) reconstructions from high-resolution serial block-face electron microscopy (SBEM) to visualize synaptic ultrastructure in layer 3 of the LPFC.

Methods

Animals and housing

Rhesus monkeys (Macaca mulatta) derived from of a larger study of normal aging in nonhuman primates conducted over the course of 14 years [2, 9, 17, 19, 23, 24, 42, 69], were used in this study. Each age group was composed of three monkeys randomly selected for ultrastructural investigations (Table 1). The cutoffs for age groups were determined based on a suggested ratio of 1:3 between rhesus monkeys and human years of age [43]. Middle age, commonly defined as being between the ages of 40 and 60 in humans, corresponds to the age range of 13–20 years of age in rhesus monkeys, young (6–12 years), and aged (21–28 years).

Table 1.

Study subjects. Age and sex of monkeys included in each study group

Age group	Sex	Age (years)
Young	Male	8.5
	Male	8.5
	Female	8.4
MAcur	Male	16.8
	Male	17.8
	Female	16.6
MActrl	Male	16.8
	Female	16.2
	Female	17
Aged	Female	27.5
	Female	26
	Male	27

All monkeys, obtained from various National Primate Research Centers and private vendors, received a medical examination before entering the study. This included serum chemistry, hematology, urine analysis, and fecal analysis, along with magnetic resonance imaging (MRI) that ensured the absence of neurological damage. Animals were individually housed under a 12-h light/dark cycle at the Animal Science Center (ASC) facility of Boston University Medical Campus, which is fully approved by the Association for Assessment and Accreditation of Laboratory Animal Care (AAALAC). The monkeys were kept in colony rooms with auditory and visual contact with other monkeys. Animals were provided with dietary items (frozen yogurt and fruit) and devices (chewing toys, foraging devices, and audio and visual stimulation) as part of enrichment procedures that meet or exceed USDA guidelines. All maintenance and research procedures were conducted according to the guidelines of the National Institutes of Health and the Institute of Laboratory Animal Resources Guide for the Care and Use of Laboratory Animals and were approved by the Boston University Institutional Animal Care and Use Committee. Feeding took place once a day and consisted of 12–20 biscuits per day (based on their weight) of Purina Monkey Chow (Purina Mills, St. Louis, MO) supplemented by fruit. Water was available continuously, and during testing, small pieces of fruit or candy were used as rewards. MA monkeys were randomly assigned to the MA curcumin subgroup (MAcur) or the MA control subgroup (MActrl). The MA group received either a daily dietary supplement of 500 mg of curcumin (MAcur) on an optimized formulation for enhanced bioavailability (Longvida, Verdure Sciences, Noblesville, IN) or a mix of dextrin and tartrazine to match the color of active curcumin (MActrl). The MA subgroups received their respective diets for a total duration of 2 years.

Tissue collection and processing

The monkeys were euthanized as described [69]. Briefly, after perfusion with ice-cold Kreb–Henseleit buffer (concentrations in mM: 6.4 Na₂HPO₄, 1.4 Na₂PO₄, 137 NaCl, 2.7 KCl, 5 glucose, 0.3 CaCl₂, 1 MgCl₂, pH 7.4), fresh tissue blocks of ~ 1 × 1 mm were dissected from LPFC area 46 and immersion-fixed in 4% paraformaldehyde in 0.1 M phosphate buffer (PB). Blocks were temporarily stored in 4% paraformaldehyde at 4 °C until further processing. We then sectioned smaller blocks at 100 µm in the coronal plane to contain all cortical layers and white matter using a vibratome (Leica VT1000S, Leica Biosystems, Deer Park, IL) and postfixed in 3% glutaraldehyde and 4% paraformaldehyde in a variable wattage microwave (Biowave Ted Pella, Redding, CA), at 150 W, 45 °C, for 5–10 min, until the sample temperature reached between 30 and 40 °C. After postfixation, all samples were thoroughly rinsed in 0.1 M sodium cacodylate buffer (3 × 5 min) and then postfixed in a solution of 2% OsO₄ (Ted Pella) with 1.5% potassium ferrocyanide (Sigma-Aldrich, St. Louis, MO) in 0.1 M sodium cacodylate buffer for 1 h, including a single microwave session at 100 W, 30 °C for 4 min (Biowave Ted Pella). After rinsing in distilled water, the samples were incubated in 1% thiocarbohydrazide for 1 h at 60 °C. A secondary postfixation was performed using 2% OsO₄ in distilled water for 1 h (microwaved at 100 W for 4 min). En-bloc staining was carried out by overnight incubation in 2–6% uranyl acetate (in distilled water; Electron Microscopy Sciences, Hatfield, PA) at room temperature. The samples were then stained with lead aspartate solution (35 mM lead nitrate, Electron Microscopy Sciences, in aspartic acid, Sigma-Aldrich) for 1 h at 60 °C. Dehydration of the sample was performed using an ascending series of ethyl alcohol (50, 75, 85, 95, 100%). Subsequently, the samples were infiltrated with 100% propylene oxide (Electron Microscopy Sciences) for 20 min, followed by overnight infiltration in a 1:1 mixture of propylene oxide and EMbed 812 resin (Electron Microscopy Sciences). The samples were then transferred to fresh 100% EMbed 812 resin for 6 h under a vacuum desiccator, flat-embedded in ACLAR film or block-embedded in BEEM capsules (Ted Pella) depending on the block-face size, and cured for 48 h at 60 °C.

Serial block-face electron microscopy, image three-dimensional reconstruction, and analysis

Blocks were first faced and sectioned at 1 µm using an ultramicrotome to identify the regions of interest (ROIs) and then imaged at Cleveland Clinic Lerner Research Institute’s 3DEM Ultrastructural Imaging and Computation Core or Boston University. The tissue was prepared as described [44]. Briefly, the serial block-face EM (SBEM) protocol for neurological tissue is automatically performed in 1 day using the programmable mPrep ASP-1000 Automated Specimen Processor (ASP) followed by resin curing SBEM. The ASP enables high-speed preparation of biological tissues with rapid and repeated fluid exchanges that accelerate reagent diffusion into specimens. 3D-SBEM images were acquired using a Zeiss Sigma VP scanning electron microscope (Carl Zeiss, White Plains, NY) equipped with a GATAN 3-view chamber for serial block-face imaging. The regions of interest (ROIs) were obtained from layer 3 of the LPFC from each block. For each ROI, 150–200 serial images were acquired at 4.5-nm/pixel x–y resolution and 65-nm z resolution (Fig. 1). We used Reconstruct software (RRID: SCR_002716; [45, 46]) to analyze SBEM stacks (15–200 sections each). To prevent potential bias, reconstructions and data analysis were performed blinded to age and treatment groups for the analysis of synapse density. An unbiased 3D grid template with four counting frames of 10 × 6 µm was applied to each series. The counting frames are composed of two exclusion and two inclusion planes. All synapses (axodendritic and shaft) within the counting frame or intersecting the inclusion planes were counted, whereas synapses outside the counting frame or intersecting any of the exclusion planes were not counted (Fig. 2). The asymmetric (excitatory) and symmetric (inhibitory) synapses were manually identified following previously defined criteria [47, 48]. We also performed a morphometric analysis of asymmetric and symmetric axospinous synapses by manually tracing presynaptic boutons, postsynaptic dendritic spines, and postsynaptic density (PSD) using the “free-hand” drawing tool in Reconstruct. Presynaptic mitochondria (mitochondria contained in the presynaptic bouton) and non-synaptic mitochondria (mitochondria outside the synaptic terminal, located in dendrites) were also traced with the same tool. All structures that were not completely within the ROI were excluded. The manual segmentations were then 3D-reconstructed to obtain the total volume and surface area of the structures.

Fig. 1.

Representative electron micrographs of the neuropil in layer 3 of the rhesus monkey LPFC. Micrographs from aged (A), middle-aged (B), and young rhesus monkeys (C). The field contains representative examples of dendrites, dendritic spines, and axon terminals forming asymmetric and symmetric synapses. The arrows label the junction between the pre- and postsynaptic components. Ad₁: asymmetric axodendritic synapse; Ad₂: symmetric axodendritic synapse; Asp₁: asymmetric axospinous synapse; Asp₂: symmetric axospinous synapse; B: presynaptic bouton; D: dendrite; M₁: synaptic mitochondria; M₂: non-synaptic mitochondria; Sp: dendritic spine; SA: spinous apparatus. Scale bars = 1 µm

Fig. 2.

Quantification and 3D reconstruction of LPFC synapses. (A) Field of view showing 2D segmentation from a single SBEM reference section. (B) The counting frame used for synapse density quantifications is shown with the inclusion (green) and exclusion limits (red). (C) Paired 3D projections from manual segmentation of asymmetric, perforated, axospinous synapses in MA monkeys. (D, F) Representative views from 2D segmentations of axospinous synapses and their paired 3D projections containing mitochondria, in aged (E) and MA (G) monkeys. Included presynaptic boutons are segmented in purple, postsynaptic spine in magenta, postsynaptic density (PSD) in green, and mitochondria in maroon. Scale bars = 1 µm

Statistics

Data were analyzed using GraphPad Prism (version 10.4.0 for Windows, GraphPad Software, Boston, MA). Associations between two variables were assessed using Pearson correlations. One-way analysis of variance (ANOVA) followed by Tukey’s honestly significant difference (HSD) post-hoc test was used to assess differences between the different groups in synaptic density and volume changes. To perform statistical comparisons of postsynaptic targets, as well as asymmetric and symmetric synapse proportions, the χ² test was used for contingency tables. To test differences in size of the different synaptic components, between age groups, we used the Wilcoxon signed-rank test and Kolmogorov–Smirnov (KS) to compare cumulative distributions. For all tests, significance was set at an α value of 0.05.

Results

Axospinous excitatory synapse density decreases in LPFC layer 3 of rhesus monkeys during aging

We observed synapse loss with age in LPFC layer 3, with a 27% decrease in total synapse density in the aged and a 25% decrease in the MA subgroups, compared to the young monkeys (one-way ANOVA, p < 0.0023; Fig. 3 A; Table 2). This loss seemed to be mainly driven by a decrease in axospinous synapses, with a 38% axospinous synapse loss in aged monkeys, a 37% decrease in MActrl, and a 33% decrease in MAcur, that were significant compared to young monkeys (oneway ANOVA, p < 0.0010; Fig. 3 A; Table 2). Analyses by synapse type revealed significant loss in asymmetric (excitatory) axospinous synapses in the aged and the MA groups compared to the young group (one-way ANOVA, p < 0.0011; Fig. 3B, C; Table 2). Within the asymmetric synapse categories, we found a significant increase in the density of perforated axospinous synapses in aged and MA groups compared to young monkeys (one-way ANOVA, p < 0.0025; Fig. 3 C; Table 2). Aged monkeys showed a significant increase in asymmetric shaft synapses (one-way ANOVA, asymmetric shaft synapses, p = 0.001) but not symmetric (inhibitory) synapses (p = 0.4866; Fig. 3D; Table 2). Despite the differences in synapse density between groups, we did not find significant changes in the chi-squared test applied to analyze the proportion of total axospinous and shaft synapses between groups (χ2, axospinous vs. shaft synapses, p = 0.188, Fig. 3B). Similarly, no significant changes in postsynaptic targets between groups (χ2, axospinous asymmetric vs. symmetric synapses, p = 0.7417; shaft asymmetric vs. symmetric synapses, p = 0.5222, Fig. 3D).

Fig. 3.

Age-related changes in synaptic density in rhesus monkey LPFC layer 3. (A) Total and axospinous synapses but not shaft synapse densities decreased with age. (B) Proportion of postsynaptic targets in each age group. (C) Axospinous symmetric and asymmetric synapse densities decreased during aging. (D) Proportion of shaft and axospinous synapses in each age group. MAcur, middle-aged with dietary curcumin; MActrl, middle-aged control. One-way ANOVA, **p < 0.01; *p < 0.05

Table 2.

Detailed statistical tests results

Figure	Statistical test	Variable measured	Comparison	Adjusted P value	Effect with Age/diet
Fig. 3	One-way ANOVA*	Total synaptic density	Young vs. Aged Young vs. MAcur Young vs. MActrl MAcur vs. MActrl MAcur vs. Aged MActrl vs. Aged	P<0.0043 P=0.0065 P<0.0043 P=0.9867 P=0.9866 P>0.9999	Decrease with age Decrease with age Decrease with age No change with diet No change with age No change with age
Fig. 3	One-way ANOVA*	Axospinous synaptic density	Young vs. Aged Young vs. MAcur Young vs. MActrl MAcur vs. MActrl MAcur vs. Aged MActrl vs. Aged	P<0.0017 P<0.0043 P<0.0020 P=0.9112 P=0.8494 P=0.9986	Decrease with age Decrease with age Decrease with age No change with diet No change with age No change with age
Fig. 3	One-way ANOVA*	Shaft synaptic density	Young vs. Aged Young vs. MAcur Young vs. MActrl MAcur vs. MActrl MAcur vs. Aged MActrl vs. Aged	P=0.0911 P=0.9845 P=0.2755 P=0.4205 P=0.1481 P=0.8480	No change with age No change with age No change with age No change with diet No change with age No change with age
Fig. 3	One-way ANOVA*	Axospinous asymmetric synaptic density	Young vs. Aged Young vs. MAcur Young vs. MActrl MAcur vs. MActrl MAcur vs. Aged MActrl vs. Aged	P=0.0013 P=0.0044 P=0.0028 P=0.9790 P=0.7270 P=0.9080	Decrease with age Decrease with age Decrease with age No change with diet No change with age No change with age
Fig. 3	One-way ANOVA*	Axospinous asymmetric perforated synaptic density	Young vs. Aged Young vs. MAcur Young vs. MActrl MAcur vs. MActrl MAcur vs. Aged MActrl vs. Aged	P=0.0358 P=0.1040 P=0.9304 P=0.2356 P=0.8714 P=0.0829	Increase with age No change with age No change with age No change with diet No change with age No change with age
Fig. 3	One-way ANOVA*	Shaft asymmetric synaptic density	Young vs. Aged Young vs. MAcur Young vs. MActrl MAcur vs. MActrl MAcur vs. Aged MActrl vs. Aged	P=0.1120 P=0.9372 P=0.5322 P=0.1195 P=0.0600 P=0.0651	No change with age No change with age No change with age No change with diet No change with age No change with age
Fig. 3	One-way ANOVA*	Axospinous symmetric synaptic density	Young vs. Aged Young vs. MAcur Young vs. MActrl MAcur vs. MActrl MAcur vs. Aged MActrl vs. Aged	P=0.0660 P=0.0605 P=0.0771 P=0.7602 P>0.9999 P=0.7270	No change with age No change with age No change with age No change with diet No change with age No change with age
Fig. 3	One-way ANOVA*	Shaft symmetric synaptic density	Young vs. Aged Young vs. MAcur Young vs. MActrl MAcur vs. MActrl MAcur vs. Aged MActrl vs. Aged	P=0.7265 P=0.9998 P=0.9505 P=0.9677 P=0.6851 P=0.4416	No change with age No change with age No change with age No change with diet No change with age No change with age
Fig. 4	Kolmogorov-Smirnov ‡	Asymmetric bouton surface	Young vs. Aged Young vs. MAcur Young vs. MActrl MAcur vs. MActrl MAcur vs. Aged MActrl vs. Aged	P=0.0007 P<0.0001 P=0.0084 P<0.0001 P=0.1422 P=0.0095	Increase with age Increase with age Increase with age Increase with diet No change with age Increase with age
Fig. 4	Kolmogorov-Smirnov ‡	Asymmetric spine surface	Young vs. Aged Young vs. MAcur Young vs. MActrl MAcur vs. MActrl MAcur vs. Aged MActrl vs. Aged	P<0.0001 P<0.0001 P<0.0001 P=0.0023 P=0.0422 P=0.8484	Increase with age Increase with age Increase with age Increase with diet Decrease with age No change with age
Fig. 4	Kolmogorov-Smirnov ‡	Asymmetric PSD surface	Young vs. Aged Young vs. MAcur Young vs. MActrl MAcur vs. MActrl MAcur vs. Aged MActrl vs. Aged	P<0.0001 P<0.0001 P<0.0001 P=0.0081 P=0.0019 P=0.1545	Increase with age Increase with age Increase with age Increase with diet Decrease with age No change with age
Fig. 4	Kolmogorov-Smirnov ‡	Symmetric bouton surface	Young vs. Aged Young vs. MAcur Young vs. MActrl MAcur vs. MActrl MAcur vs. Aged MActrl vs. Aged	P=0.9323 P=0.1156 P=0.2601 P=0.4396 P=0.4554 P=0.3460	No change with age No change with age No change with age No change with diet No change with age No change with age
Fig. 4	Kolmogorov-Smirnov ‡	Symmetric spine surface	Young vs. Aged Young vs. MAcur Young vs. MActrl MAcur vs. MActrl MAcur vs. Aged MActrl vs. Aged	P=0.0032 P<0.0001 P=0.0627 P=0.0191 P=0.0763 P=0.7499	Increase with age Increase with age No change with age Increase with diet No change with age No change with age
Fig. 4	Kolmogorov-Smirnov ‡	Symmetric PSD surface	Young vs. Aged Young vs. MAcur Young vs. MActrl MAcur vs. MActrl MAcur vs. Aged MActrl vs. Aged	P<0.0001 P=0.0055 P=0.4995 P=0.2459 P=0.0986 P=0.8465	Increase with age Increase with age No change with age No change with diet No change with age No change with age
Fig. 5	One-way ANOVA*	Bouton volume	Young vs. Aged Young vs. MAcur Young vs. MActrl MAcur vs. MActrl MAcur vs. Aged MActrl vs. Aged	P=0.0073 P=0.3001 P=0.0819 P=0.0002 P=0.5193 P<0.0001	Increase with age No change with age No change with age Increase with diet No change with age Increased with age
Fig. 5	One-way ANOVA*	Spine volume	Young vs. Aged Young vs. MAcur Young vs. MActrl MAcur vs. MActrl MAcur vs. Aged MActrl vs. Aged	P=0.0004 P<0.0001 P<0.0001 P<0.0001 P=0.0021 P=0.9987	Increase with age Increase with age Increase with diet Increase with diet Increase with age No change with age
Fig. 5	One-way ANOVA*	PSD surface	Young vs. Aged Young vs. MAcur Young vs. MActrl MAcur vs. MActrl MAcur vs. Aged MActrl vs. Aged	P<0.0001 P<0.0002 P=0.0003 P=0.1147 P=0.9996 P=0.2460	Increase with age Increase with age Increase with age No change with diet No change with age No change with age
Fig. 6	Kolmogorov-Smirnov ‡	Asymmetric mitochondria surface	Young vs. Aged Young vs. MAcur Young vs. MActrl MAcur vs. MActrl MAcur vs. Aged MActrl vs. Aged	P<0.0001 P<0.0001 P=0.0018 P=0.0002 P=0.1147 P=0.1267	Increase with age Increase with age Increase with age Increase with diet No change with age No change with age
Fig. 6	Kolmogorov-Smirnov ‡	Symmetric mitochondria surface	Young vs. Aged Young vs. MAcur Young vs. MActrl MAcur vs. MActrl MAcur vs. Aged MActrl vs. Aged	P=0.2999 P=0.3520 P=0.9164 P=0.3510 P>0.9999 P=0.5237	No change with age No change with age No change with age No change with diet No change with age No change with age
Fig. 6	One-way ANOVA*	Synaptic mitochondria volume	Young vs. Aged Young vs. MAcur Young vs. MActrl MAcur vs. MActrl MAcur vs. Aged MActrl vs. Aged	P<0.0001 P=0.02532 P<0.0001 P<0.0001 P=0.3728 P=0.0182	Increase with age Increase with age Increase with age Increase with diet No change with age No change with age
Fig. 6	One-way ANOVA*	Synaptic mitochondria surface area	Young vs. Aged Young vs. MAcur Young vs. MActrl MAcur vs. MActrl MAcur vs. Aged MActrl vs. Aged	P<0.0001 P=0.0185 P<0.0001 P=0.0005 P=0.3272 P=0.2408	Increase with age Increase with age Increase with age Increase with diet No change with age No change with age
Fig. 6	One-way ANOVA*	Synaptic mitochondria length	Young vs. Aged Young vs. MAcur Young vs. MActrl MAcur vs. MActrl MAcur vs. Aged MActrl vs. Aged	P<0.0001 P<0.0001 P<0.0001 P<0.0001 P=0.0002 P>0.9999	Increase with age Increase with age Increase with age Increase with diet No change with age No change with age
Fig. 6	One-way ANOVA*	Non-synaptic mitochondria volume	Young vs. Aged Young vs. MAcur Young vs. MActrl MAcur vs. MActrl MAcur vs. Aged MActrl vs. Aged	P<0.0001 P=0.0007 P<0.9999 P=0.9516 P=0.0003 P=0.0020	Decrease with age Increase with age Increase with age No difference with diet Decrease with age Decrease with age
Fig. 6	One-way ANOVA*	Non-synaptic mitochondria surface area	Young vs. Aged Young vs. MAcur Young vs. MActrl MAcur vs. MActrl MAcur vs. Aged MActrl vs. Aged	P=0.5545 P<0.0001 P<0.0001 P=0.9996 P<0.0001 P<0.0001	No change with age Increase with age Increase with age No difference with diet Decrease with age Decrease with age
Fig. 6	One-way ANOVA*	Non-synaptic mitochondria length	Young vs. Aged Young vs. MAcur Young vs. MActrl MAcur vs. MActrl MAcur vs. Aged MActrl vs. Aged	P=0.2661 P<0.0001 P<0.0001 P=0.5897 P<0.0001 P<0.0001	No change with age Increase with age Increase with age No difference with diet Decrease with age Decrease with age

*Results from the one-way ANOVA followed by Tukey’s honestly significant difference (HSD) post-hoc test. Adjusted p values are shown. Statistical significance was set at p ≤ 0.05

‡Results from the Kolmogorov–Smirnov test to compare cumulative distributions. Statistical significance was set at p ≤ 0.05

Ultrastructural features and pre- and postsynaptic relationships in LPFC layer 3 of rhesus monkeys are dependent on age and diet

We first investigated whether aging altered the size distributions of the different synaptic components by synapse type within each group to detect any changes with age or the effects of dietary curcumin in the MA cohort. Overall, asymmetric synapses showed greater differences between groups in the Wilcoxon signed-rank test (p < 0.0001) and the KS test (Table 2) in surface area distributions compared to symmetric synapses (Fig. 4). Both MA groups had a significant rightward shift (toward larger sizes) in bouton, spine, and PSD surface area distributions of asymmetric synapses compared to the young group (Fig. 4A–C; Table 2). The aged group showed larger sizes in boutons of symmetric but not asymmetric synapses compared to young monkeys (Fig. 4 A, D; Table 2). Comparisons within the MA group showed an effect of dietary curcumin on asymmetric synapse and spine size. Specifically, MAcur exhibited larger size of asymmetric boutons and PSDs, as well as the larger spine head size for total spines and the subset that also received symmetric synapses (Fig. 4A–C, E; Table 2). In addition, MAcur exhibited larger spine and PSD sizes than aged monkeys (Fig. 4B, C; Table 2).

Fig. 4.

Cumulative distributions of the surface area of different synaptic components in axospinous synapses in rhesus monkey LPFC layer 3. Axospinous asymmetric synapse size distributions (A–C) are affected more severely by age and after dietary curcumin supplementation than symmetric synapses (D–F). Wilcoxon Rank Sum test p < 0.001 for all groups. Kolmogorov–Smirnov test (Table 2). MAcur, middle-aged with dietary curcumin; MActrl, middle-aged control

We also examined the differences in the mean volume of pre-and postsynaptic components, as well as the average area of the PSD surface, and studied their relationship within each group (Fig. 5 A, B; Table 2). Bouton volume increased in aged monkeys compared to young animals and MActrl, with no significant differences in bouton size with the MAcur monkeys (one-way ANOVA, p < 0.0001; Fig. 5 A; Table 2). Average spine volumes were larger in aged and both MA subgroups compared to young, with MAcur showing the largest spine volume compared to all other groups (one-way ANOVA, p < 0.0001; Fig. 5B; Table 2). We found that PSD was larger in both MA and aged groups, in comparison with young monkeys (one-way ANOVA, p < 0.0001; Fig. 5 C; Table 2). Based on studies showing the importance of size relationships of the different synaptic elements [49–51], we analyzed the size relation of the different pre- and postsynaptic components with each group. A positive correlation of bouton volume and spine volume was observed within each group, that was slightly stronger in aged and MA monkeys in comparison to young (Pearson’s correlation, young: r = 0.4430, p < 0.0001; MAcur: r = 0.5354, p < 0.0001; MActrl: r = 0.5320, p < 0.0001; aged: r = 0.5522 p < 0.0001; Fig. 5D). However, when we analyzed the correlation between postsynaptic spine volume and PSD surface area, we found that the aged group presented a significantly weaker correlation than the MA and young groups, whereas the MAcur subgroup showed the highest correlation coefficient among all other groups (Pearson’s correlation, young: r = 0.8289, p < 0.0001; MAcur: r = 0.8419, p < 0.0001; MActrl: r = 0.7839, p < 0.0001; aged: r = 0.4121, p < 0.0001; Fig. 5E).

Fig. 5.

Ultrastructural analysis of 3D reconstructed axospinous synaptic components in rhesus monkey LPFC layer 3. The average volume of presynaptic boutons (A), volume of postsynaptic dendritic spines (B), and PSD surface area (C) of axospinous synapses changed with age and curcumin treatment. One-way ANOVA, ***p < 0.001; **p < 0.01; *p < 0.05. Size correlation between pre- and postsynaptic components (D) and PSD and spine size (E) of axospinous synapses. The ultrastructure of the different synaptic components in axospinous synapses changed proportionally in middle aged, but not in aged monkeys, affecting the size relationship between synaptic components. Pearson’s correlation (Table 2). MAcur, middle-aged with dietary curcumin; MActrl, middle-aged control

Morphometric changes in synaptic and non-synaptic mitochondria in LPFC layer 3 associated with age and diet.

Given that mitochondrial size correlates with energy production [52, 53] and that mitochondria are implicated in age-related increase of oxidative stress [33, 34, 54], we analyzed the size of mitochondria within the presynaptic terminal, as well as mitochondria located outside synapses, to detect any differences with age or curcumin administration. We implemented mitochondria morphometric analysis by quantifying the volume, surface area, and length. In non-synaptic mitochondria (in dendrites outside the synaptic terminal), the volume was larger for both MA groups, but not aged individuals compared to young (one-way ANOVA, volume: p < 0.0001; Fig. 6 A; Table 2). The MA groups showed the larger surface area and length of non-synaptic mitochondria compared to aged and young monkeys (one-way ANOVA surface area: p < 0.0001; length: p < 0.0001; Fig. 6B, C; Table 2). Our results showed a significant increase in the average volume, surface area, and length of presynaptic mitochondria in MA and aged groups compared to the young (one-way ANOVA, volume: p < 0.0001; surface area: p < 0.0001; length: p < 0.0001; Fig. 6D–F; Table 2). The MAcur group showed significantly larger volume and surface than MActrl and young groups (Fig. 6D, E; Table 2) and longer synaptic mitochondria than all other groups (Fig. 6 F; Table 2). In the size distributions by synapse type, the Wilcoxon signed-rank test (p < 0.0001) and Kolmogorov–Smirnov test (Table 2) showed that mitochondria contained in asymmetric presynaptic terminals were significantly larger for MAcur compared to all other groups (Fig. 6G; Table 2). However, we found that presynaptic mitochondria contained in symmetric synaptic boutons did not show significant size differences between groups (Fig. 6H; Table 2). Representative synaptic and non-synaptic mitochondria sampled in the LPFC are shown in Fig. 7. The non-synaptic mitochondria analyzed were straight and c-shaped, whereas synaptic mitochondria were also found in toroidal (doughnut-shaped).

Fig. 6.

Age-related changes andeffect ofcurcumin on different morphometric parameters in synaptic and non-synaptic mitochondria in rhesus monkey LPFC layer 3.(A–C) Non-synaptic mitochondrial volume, surface area, and length. (D–F) Synaptic mitochondrial volume and surface area. The cumulative distributions of mitochondrial surface area located in asymmetric synaptic terminals (G) was more affected by age and dietary curcumin than mitochondria within symmetric (H) presynaptic boutons. MAcur, middle-aged with dietary curcumin; MActrl, middle-aged control. One-way ANOVA, ***p < 0.001; **p < 0.01; *p < 0.05

Fig. 7.

3D reconstruction of synaptic and non-synaptic mitochondria morphologies in LPFC. Two-dimensional segmentations and paired orthogonal projections of representative synaptic and non-synaptic mitochondria. Scale bars = 1 µm

Discussion

We used three-dimensional reconstructions from SBEM for an in-depth analysis of synaptic density and changes in the ultrastructure of synapses and mitochondria in layer 3 of area 46 in the LPFC of rhesus monkeys across the adult lifespan. Our data revealed notable age-related differences in synapse density and morphology of the synaptic components, as well as changes in synaptic and non-synaptic mitochondria that might contribute to cognitive decline during aging. We also observed some important differences in axospinous synapses and mitochondrial morphology in the middle-aged monkeys treated with chronic dietary curcumin (MAcur group), which is consistent with a potential neuroprotective effect of curcumin in this age group.

Age-related changes and dietary curcumin effects on synaptic density in LPFC

We found a decrease in total synapse density in layer 3 of area 46 in MA and aged groups, compared with young monkeys, driven by the loss of asymmetric axospinous synapses. These results agree with reports in the LPFC of humans as well as macaque and marmoset monkeys during aging that describe similar synapse density changes in the cerebral cortex with age accompanied by a loss of dendritic spines and a decrease in the arborization of dendritic trees on pyramidal neurons in layers 2 and 3 in areas 46 and 10 of the LPFC [6, 7, 9–11, 14, 55–59]. Other important plasticity changes in cortical neurons suggested to affect synaptic function are synapse size and the proportion and distribution of the different postsynaptic targets [60–62]. EM studies in the human and NHP cortical neuropil have reported a larger proportion of asymmetric (excitatory) in comparison to symmetric (inhibitory) synapses, which seems to remain constant with age [9, 63, 64]. Our data confirmed this larger proportion of asymmetric synapses in all age groups analyzed. Also, like previous EM studies in the same cortical area in macaques [9], our data revealed that, in LPFC layer 3, the proportion of postsynaptic targets (i.e., axospinous and shaft synapses) does not change significantly during aging.

We also assessed whether there were any differences in synaptic density between the middle-aged groups resulting from chronic administration of dietary curcumin. Curcumin is a polyphenol [38] that has antioxidant and anti-inflammatory properties [35–37] and has been demonstrated to slow or reverse oxidative stress in the brain during aging in some studies [39, 40]. The efficacy of chronic dietary curcumin administration in improving spatial working memory in MA rhesus monkeys was recently demonstrated by our group [23, 42]. The present data did not show any significant effect of curcumin on total synapse numbers, indicating that curcumin does not ameliorate cognitive decline by prevention of synapse loss. This finding is also consistent with our previous report that dietary curcumin does not prevent significant spine loss seen in MA rhesus monkey prefrontal cortex [23].

Age-related changes and dietary curcumin effects on the ultrastructure of axospinous synapses in LPFC

We further explored the ultrastructural features of axospinous synapses across groups, as alterations in these may reflect changes in plasticity during aging or in response to curcumin administration. Changes in dendritic spine morphology associated with age-related working memory impairment have been amply described in NHP [9, 11, 12, 14, 57, 59, 65–67]. To discover subtle differences in age-related size changes in synaptic components, we first examined the size distributions of spines, PSDs, and boutons by synapse type across all groups. Overall, our data confirmed that the synaptic elements in asymmetric synapses presented larger sizes and experienced more dramatic differences in size with age than symmetric synapses. The cumulative size distributions showed a larger surface area of postsynaptic spines and PSD in asymmetric axospinous synapses in both MA groups compared to young and aged monkeys. Between the MA groups, the asymmetric synapses of MAcur monkeys had larger spine, bouton, and PSD sizes than MActrl. In addition, spines that specifically received symmetric synapses, in addition to asymmetric synapses, were larger in MAcur compared to MActrl. Interestingly, aged individuals showed larger boutons, spines, and synaptic area, specifically associated with symmetric synapses compared to young monkeys. This parallels previous evidence showing that axodendritic and axosomatic terminals forming symmetric (inhibitory) synapses in layers 2–3 of area 46 in aged macaques become larger with age [68]. We suggest that a similar compensatory homeostatic increase in axospinous symmetric synapse size occurs in response to the potential excitatory-inhibitory imbalance driven by age-related synapse loss. Our study revealed that changes in the morphology of pre- and postsynaptic elements occur in parallel to synapse loss in LPFC layer 3 during aging. Mounting evidence suggests a strong correlation between synapse size and synaptic strength and efficacy in high-order association regions of the neocortex [62, 64, 69–71]. However, a previous study describing synapse loss during aging in macaque LPFC did not find differences in the length of the profile of synaptic junctions with age [9]. Here, we used 3D reconstructed serial images obtained by SBEM that allowed us to measure the synaptic surface area in the reconstructed synapses. We found that synaptic apposition surface area did increase with age and assessed whether the pre- and postsynaptic elements scale proportionally to maintain synaptic function during aging. The ultrastructural size principle has been used to describe the importance of a strong correlation between the different components of the synapse. It is hypothesized that the violation of this principle undermines synaptic efficacy and might affect cognitive function [49–51, 67]. Violations of this principle might entail that the different synaptic components do not change proportionally to meet the energy demands of synapses and maintain effective synaptic transmission [49–51, 67]. To test this hypothesis, we analyzed the correlations between the sizes of the synaptic elements. We found evidence that might point to differences in pre- and post-synaptic communication in the aged monkeys. Our data show that bouton, but not spine or PSD, increased in size in aged monkeys compared to all other groups. Additionally, the postsynaptic spine did not scale properly with synapse size, impacting the correlation between spine and PSD surface, which was almost half in the aged group when compared to young and MA monkeys. Altogether, our data point to an improper synaptic scaling that might take place after middle-age. We studied the effects of curcumin at this stage. When we analyzed the differences in the morphology of the different synaptic components between the MA groups, our data showed a better preservation of the ultrastructure of axospinous synapses in the MAcur group, with a larger spine and bouton sizes and also a stronger size relationship between spine and PSD. These effects on the preservation of the synapse morphology by curcumin might in part explain the different changes between groups on neuronal hyperexcitability observed in the prefrontal cortex [23], and the delaying of age-related working memory decline by curcumin when compared with MActrl group [42].

Age-related changes and dietary curcumin effects on synaptic and non-synaptic mitochondria in LPFC

In addition to changes in the ultrastructure of pre- and postsynaptic components, we also observed age-related changes in the morphology of presynaptic mitochondria and extended the study to non-synaptic mitochondria located in dendrites, outside the synaptic terminal. Mitochondria contained in presynaptic boutons were much shorter than non-synaptic mitochondria. Synaptic mitochondria were longer and larger (in both volume and surface area) in MA and aged monkeys, in comparison to young, whereas the same morphometric parameters for non-synaptic mitochondria increased in the MA groups but decreased in aged monkeys. This decrease in size and length observed in aged monkeys compared to MA might impact mitochondrial metabolism and energy supply during aging. Mitochondria help maintain synaptic function and plasticity. Whereas non-synaptic mitochondria are distributed throughout the dendrite, outside active synapses, synaptic mitochondria are located at the synaptic terminal, where they also play an important role in glutamate synthesis and metabolism directly [72–75]. We suggest that the differences in synaptic and non-synaptic mitochondria morphology between groups observed during aging might have different physiological significance, considering their specific roles based on location [28, 31]. In addition, mitochondria located at synapses present more functional and proteomic heterogeneity than non-synaptic mitochondria because they are subject to more dynamic environmental changes at the synaptic compartment [28, 31]. Synaptic mitochondria have a longer lifespan, which renders them more vulnerable to increases in oxidative stress or calcium insults than non-synaptic mitochondria [28, 31, 76, 77]. Accumulated insults during aging interfere with the regulation of the reduction–oxidation homeostasis and lead to a decline in mitochondrial function that has been directly attributed to morphological changes such as length, volume, or surface area [33, 34, 54]. We identified changes in mitochondrial morphology with aging that also differed depending on their subcellular location, supporting previous evidence [33, 34, 54] that these age-related changes in morphology affect synaptic and non-synaptic mitochondrial function. One important additional observation was that the MAcur group showed larger mitochondrial volumes and surface areas compared to all other groups, as well as longer synaptic mitochondria than aged and MA control monkeys. The beneficial effects of curcumin on mitochondrial morphology may be related to its antioxidative properties that may exert a protective effect against abnormal increases in oxidative stress during aging [33, 34, 54].

Conclusion

This study confirmed that there is a reduction in axospinous synapse density and revealed details of alterations in the ultrastructure of synaptic components in layer 3 of the LPFC during aging in rhesus monkeys. We observed a loss of axospinous synapses in both MA and aged monkeys. However, in the analysis of the ultrastructure of synapses in the MA, unlike in aged monkeys, the different synaptic components showed a parallel increase in size and stronger positive correlations with each other. In addition, the size and morphology of synaptic (located in presynaptic boutons) and non-synaptic mitochondria (located in dendrites) seem to proportionally increase with age until MA to then decline in aged monkeys. While curcumin diet does not reduce synapse or spine loss [23], the present study shows that long-term curcumin supplement starting at middle age, might help to prevent age-related morphological changes to synapse and mitochondria morphology that may underlie working memory changes associated with aging in some subjects [23, 40]. MAcur shows a compensatory homeostatic increase in axospinous symmetric synapse size, along with a stronger size relationship between spine and PSD and better preservation of mitochondrial function compared to MActrl. We suggest that the antioxidative effect of curcumin might favor a better adaptational response to an increase in neuronal hyperexcitability [23] observed in the prefrontal cortex caused by synapse loss during aging. Overall, protective effects of curcumin on the ultrastructure of synaptic components and synaptic and non-synaptic mitochondria suggest a plausible mechanism by which an early intervention with a nutraceutical supplement such as curcumin may act to ameliorate cognitive decline.

Author contribution

Carmen Freire-Cobo: conceptualization, formal analysis, acquisition, visualization, writing—original draft; Maria Medalla: conceptualization, methodology, writing—review and editing; Merina Varghese: conceptualization, writing—review and editing; William G.M. Janssen: methodology, resources; Jennifer I. Luebke: conceptualization, supervision, funding, writing—review and editing; Patrick R. Hof: conceptualization, supervision, funding, writing—review and editing

Funding

This work was supported by the National Institutes of Health R01 AG059028.

Declarations

Competing interests

The authors declare no competing interests.