Neuronal vulnerability to brain aging and neurodegeneration in cognitively impaired marmoset monkeys (Callithrix jacchus)

Abstract

The investigation of neurobiological and neuropathological changes that affect synaptic integrity and function with aging is key to understanding why the aging brain is vulnerable to Alzheimer’s disease. We investigated the cellular characteristics in the cerebral cortex of behaviorally characterized marmosets, based on their trajectories of cognitive learning as they transitioned to old age. We found increased astrogliosis, increased phagocytic activity of microglial cells and differences in resting and reactive microglial cell phenotypes in cognitively impaired compared to non-impaired marmosets. Differences in amyloid beta deposition were not related to cognitive trajectory. However, we found age-related changes in density and morphology of dendritic spines in pyramidal neurons of layer 3 in the dorsolateral prefrontal cortex and the CA1 field of the hippocampus between cohorts. Overall, our data suggest that an accelerated aging process, accompanied by neurodegeneration, that takes place in cognitively impaired aged marmosets and affects the plasticity of dendritic spines in cortical areas involved in cognition and points to mechanisms of neuronal vulnerability to aging.

1-. Introduction

The study of the aging brain in nonhuman primates (NHP) is critical to investigate the age-related neurobiological and neuropathological changes affecting brain synaptic integrity and function, thus increasing vulnerability to Alzheimer’s disease (AD-like pathology. The common marmoset (Callithrix jacchus) is a small NHP with a relatively short life expectancy (~10-12 years; (Nishijima et al., 2012; Ross & Salmon, 2019). Marmosets show highly developed social and cognitive abilities (Miller, 2017; Miller et al., 2016; Nakamura et al., 2018; Nummela et al., 2019; Spinelli et al., 2004) and possess a brain architecture typical of anthropoid, with a well developed prefrontal cortex (PFC). As such, the marmoset has emerged as an ideal model for translational neuroscience research (Abbott et al., 2003; Rothwell et al., 2021) particularly for aging studies (Ross et al., 2019; Ross & Salmon, 2019; Rothwell et al., 2021; Tardif et al., 2011). As reported in other NHPs (Lacreuse et al., 2020; Ross et al., 2019; Ross & Salmon, 2019) cross-sectional studies using marmosets have demonstrated age differences in a range of cognitive domains, including executive function (Munger et al., 2017; Sadoun et al., 2019) working memory (Sadoun et al., 2019), and detoured reaching (Phillips et al., 2019). In addition, recent longitudinal studies have pinpointed the onset of age-related cognitive decline around 8 years in this species (Rothwell et al., 2022).

Few studies have documented age-related plasticity changes and neuropathology in the marmoset brain (Freire-Cobo et al., 2021). Some of the hallmarks of AD-type pathology have been described in the marmoset (Geula et al., 2002; Maclean et al., 2000; Philippens et al., 2017; Ridley et al., 2006; Rodriguez-Callejas et al., 2016), principally focusing on amyloid beta (Aβ) protein deposition in the neocortex (Geula et al., 2002; Rodriguez-Callejas et al., 2016) and the mechanism of amyloidosis (Maclean et al., 2000; Ridley et al., 2006). The involvement of glial cells was also investigated (Rodríguez-Callejas et al., 2019; Rodriguez-Callejas et al., 2016), revealing increased number of astrocytes along with changes in microglia phenotypes in the hippocampus of old marmosets (Rodríguez-Callejas et al., 2019). Other pathological changes include degeneration of white matter in the corpus callosum (Philippens et al., 2017) and tau protein hyperphosphorylation (Geula et al., 2002; Rodriguez-Callejas et al., 2016; Sharma et al., 2019). However, the relationships between brain pathology and cognitive function have not been clearly established in this species. Furthermore, individual variability in the susceptibility or resistance to AD-like neuropathology can only be understood with longitudinal studies (Arnsten et al., 2021; Rothwell et al., 2021).

Loss of dendritic spines or changes in their morphology has not been assessed in marmosets either. These structural changes likely mediate age-related cognitive decline in other NHPs, especially in macaques (Dickstein et al., 2013; Duan et al., 2003; Dumitriu et al., 2010; Motley et al., 2018; Page et al., 2002). The dorsolateral prefrontal cortex (dlPFC) and hippocampus are cortical regions involved in cognitive functions and known to be especially vulnerable to aging and AD across species (Burke & Barnes, 2010; Gallagher & Nicolle, 1993; Morrison & Hof, 2007). Cognitive functions dependent on the dlPFC, such as executive functions, working memory, and goal-directed behavior, are particularly vulnerable to aging in macaques (Moore et al., 2006; Wang et al., 2011). The hippocampus has a critical role in episodic memory (Morrison et al., 1982; Tulving & Markowitsch, 1998). Cellular and synaptic changes in the hippocampal formation during aging, may underlie general impairments in hippocampus-dependent memory abilities (Baxter & Murray, 2001; Burke & Barnes, 2010; Gallagher & Nicolle, 1993). A significant decrease in spine density of layer 3 pyramidal neurons in area 46 of the dlPFC has been reported in aged macaque monkeys (Duan et al., 2003; Dumitriu et al., 2010) and in area 7a of behaviorally impaired macaques (Motley et al., 2018). There is no current evidence describing changes in dendritic spines during aging in marmosets, and only dendritic growth and normal spine formation and pruning from birth until 2.5 years of age have been documented in the PFC of young marmosets (Ichinohe, 2015; Sasaki, 2015).

In this study, we investigate the cellular characteristics in the cerebral cortex of adult male and female marmosets as they transitioned to old age. The analyses focused on two cortical regions relevant to cognitive functions and involved pathological markers, along with a detailed stereological study of dendritic spine density and morphology in behaviorally characterized marmosets.

2-. Materials and Methods

2.1-. Subjects, cognitive testing, and cognitive aging categorization

Adult common marmosets (N=17; 9 females, 8 males) were studied for cognitive aging and neuropathology. At the start of cognitive testing, the average age of females was 4.88 years old (range: 4.21 to 6.05 years) and the average age of males was 5.12 years old (range: 3.96 to 6.86 years). Cognitive testing took place over four years and spanned the transition from middle to early old age for marmosets. The marmosets in the current study were part of a larger longitudinal aging The marmosets in the current study were part of a larger longitudinal aging study (N=28) which focused on parameters of behavioral aging (Rothwell et al., 2022). Marmosets were tested for cognitive abilities using a reversal-learning task across the four years (see Rothwell et al., 2022 for details). Reversal learning is comprised of two testing phases, simple discrimination (SD), and simple reversal (SR), using a pair of visual stimuli. Cognitive tests were administered using a touchscreen interface (Cambridge Neuropsychological Automated Battery, CANTAB) or a manual version (modified Wisconsin General Testing Apparatus, WGTA) and overall cognitive performance did not differ between these two tasks (Rothwell et al., 2022). Cognitive performance was measured by trials to criterion (criterion = 90% on 40 consecutive trials) in each testing phase. Once a marmoset reached the criterion for SD, he or she began the SR phase for that stimulus pair. When a marmoset reached the criterion on SR, he or she began a new stimulus pair. Marmosets completed SD and SR testing for three pairs of stimuli per year across four years, giving up to 12 scores for cognitive performance each for discrimination and reversal testing. In order to evaluate group level change in cognitive performance across aging, we used multilevel mixed effects models which nested repeated observations within individual marmosets (Raudenbush, 2002). These growth models were estimated using full information maximum likelihood in order to include marmosets with missing or partial data. Linear or curvilinear (i.e. quadratic) baseline growth models were fit for SD and SR testing, and the quadratic model was retained only if it provided a significantly better fit compared to the linear model (χ² model comparison test, p < 0.05). To categorize individual marmosets as cognitively impaired or not, we first visualized change in cognitive performance across aging by adding an Ordinary Least Squares (OLS) regression line to each marmoset’s trials to criterion scores for SD and SR tests. These regression lines enabled us to see which marmosets showed any sign of a declining trajectory of performance, which was indicated by positive slope (i.e. marmoset took more trials to reach criterion on subsequent tests). Marmosets with evidence of cognitive decline on either SD or SR testing were categorized as cognitively impaired across aging. A marmoset was categorized as cognitively non-impaired if he or she showed no positive slope on either SD or SR, suggesting stable or improved performance on cognitive tests across aging. Finally, we evaluated the validity of these categories by adding cognitive category (impaired, non-impaired) as a predictor variable to the baseline growth models to see if our categories significantly predicted variability around the average growth trajectories (SD & SR slopes). We also added age at start of cognitive testing as a predictor variable to assess whether age significantly predicted initial performance (SD & SR intercepts) or growth trajectories (SD & SR slopes).

2.2-. Brain perfusion and preparation

After the completion of cognitive testing, subjects were sacrificed for collection of brain tissue. The average age at death of females was 8.50 years old (range: 7.92 to 9.74) and of males was 8.71 years old (range: 7.50 to 10.33 years). Monkeys were deeply anesthetized with sodium pentobarbital (100 mg/kg, IM) and subject to transcardial perfusion with cold 4% paraformaldehyde/0.125% glutaraldehyde in phosphate buffer saline (PBS) for 15 min. Brains were removed, and bisected along the midline, postfixed in the same fixative for 48 hours, and transferred to PBS. The marmoset brains were dissected to include the regions of interest (figure 1), the CA1 field of the hippocampus, and area 8b/9 of the dlPFC in the sampled blocks (Atapour et al., 2019; Theodoni et al., 2021). Using a vibratome (VT1000S, Leica Microsystems, Buffalo Grove, IL) slices of different thickness were collected for different experimental purposes. In an iterative manner, two thick slabs (200 μm) were obtained for cell loading and 3D reconstruction, followed by a series of 4 thinner sections (50 μm) for immunohistochemical experiments. Alternating tissue collection in that way can ensure sufficient materials for both analyses and maintains a systematic-random sampling for stereology that encompasses the entire regions of interest.

Figure 1. Sampling of regions of interest in the marmoset brain.

(A) Lateral view of a marmoset brain showing the strategy for dissection of brain blocks. The vertical dashed lines mark the dissection limits for the collection of the blocks that contain the regions of interest: area 8b/9 of the PFC (A1) and CA1 of the hippocampus (A2). (B-C) Representative slabs from block A1 (B) and block A2 (C) containing the PFC and CA1, respectively (asterisks). The schematic representations at left of the slabs show the extent of the regions of interest. (D-E) Representative coronal sections stained with DAPI, showing the cellular layers in PFC (D) and hippocampal CA1 (E). The white overlays display templates (bounded grids made of stitched images) used during confocal microscopy imaging to sample the regions of interest, in layer 3 of area 8b/9 (D) and CA1 pyramidal layer (E). Rad: stratum radiale; or: stratum oriens; pyr: stratum pyramidale. D: dorsal; V: ventral; L: lateral; M: medial. Scale bars = 1 cm (A), and 100 μm (D,E).

2.3-. Immunofluorescence

Free-floating brain sections from the hippocampus and dlPFC were permeabilized in PBS and 0.2% Triton X-100 (PBST), followed by blocking with 5% normal goat serum in PBST. The incubation with the respective primary antibodies in 2% normal goat serum in PBST was performed overnight at 4°C. The following primary antibodies and dilutions were used: anti-ionized calcium-binding adaptor molecule, Iba-1 (1:500, rabbit polyclonal, Wako Chemicals, Richmond, VA, cat#019-19741, RRID:AB_839504) to detect microglia; anti-cluster of differentiation 68 antibody, CD68 (1:250, IgG1 mouse monoclonal, Thermofisher, Waltham, MA, Cat# MA5-13324, RRID:AB_10987212) to detect microglial lysosomes; anti amyloid beta antibody, MOAB-2 (1:500, mouse monoclonal IgG1, Abcam, Waltham, MA, Cat# ab11132, RRID:AB_297770) to detect Aβ deposits; anti-glial fibrillary acidic protein, GFAP (1:500, rabbit polyclonal, Agilent, Santa Clara, CA, Cat# Z0334, RRID:AB_10013382) to detect astrocytes; and phosphorylated-tau, AT8 (1:250, mouse monoclonal IgG1,Thermo Fisher Scientific, Waltham, MA, Cat# MN1020, RRID:AB_223647) to detect phosphorylated tau protein. Sections were subsequently washed in PBS and incubated with the corresponding Alexa Fluor^™ conjugated secondary antibody ( 1:5000, Thermofisher, Waltham, MA) in 2% normal goat serum in PBST for 1 h at room temperature. Sections processed in the absence of the primary antibody were used as controls for the polyclonal antibodies. For the monoclonal antibodies ( MOAB-2, AT8, CD68), control sections were incubated in the presence of IgG1 instead of the primary antibody. After serial washes in PBS, lipofuscin autofluorescence was quenched by a 30-s treatment with TrueBlack (Biotium, Fremont, CA), followed by PBS washes. Sections were then mounted and cover-slipped using VectaShield anti-fade mounting medium (Vector Laboratories, Newark, CA) containing 4’,6-diamidino-2- phenylindole (DAPI, Sigma-Aldrich, St. Louis, MO), and dried overnight.

The imaging of the different brain sections was performed using a Zeiss LSM 780 upright confocal microscope (Carl Zeiss, Jena, Germany). Three brain slices were imaged per animal and brain area, a predefined 4-6 tile template for imaging was systematically applied to each slide in the confocal microscope, to sample the entire area of interest (layer 3 of area 8b/9 in dlPFC and CA1), with a total collection of 12-18 images per region of interest per animal. Confocal 20-μm z-stacks were acquired at 1-μm intervals, using either a plan-apochromat 20x10.8 objective (MOAB-2, GFAP); 40×/1.3 oil-immersion objective and 0.6 zoom factor (Iba-1, CD68); or a 63x/1.4 NA oil-immersion objective (Iba-1, microglia morphology). Image analysis was performed using the Image J software (National Institutes of Health, Bethesda, MA) (Schneider et al., 2012). A binary image was created from the raw single-channel from the image projections of the 20 μm-thick z-stacks, followed by automated thresholding to determine the percentage of area covered with specific staining of Aβ marker in every region of interest from hippocampus CA1 and area 8b/9. Similarly, in the same regions of interest, image projections of Iba-1 and CD68 staining were used to calculate the percentage area covered by the CD68 inflammation marker within iba-1 expressing cells, relative to the total area covered by the microglial marker Iba-1 (Ayata et al., 2018). The number of astrocytes (GFAP-immunoreactive cells) and microglial (Iba-1-immunoreactive) cells with different morphological state were scored per unit area in each slide and brain area for each behavioral cohort (number of cells/number of images × single image area 0.045 mm2) using the cell counter plugin (Image J, NIH, Bethesda, MA). Microglia with different cellular and morphological characteristics were further classified based on previous descriptions (Rodriguez-Callejas et al., 2016; Streit, 2004) into resting (displaying a slightly ramified morphology and small rounded soma), activated or intermediate (hypertrophic soma and ramified cells with extensively thick and branched processes), and dystrophic cells (loss of fine branches, presence of shortened tortuous processes and/or cytoplasmic fragmentation). For a detailed morphological analysis of microglial cells, 3D reconstruction of 63x z-stacks from individual microglia were used.

Maximum intensity projections of the Iba-1 staining were converted to a binary image and only cells presenting the entire cell body were isolated (n=20 microglia/animal/region of interest). Cell morphology was analyzed using the AnalyzeSkeleton plugin (ImageJ, NIH) (Arganda-Carreras et al., 2010). The size and complexity of branch structure in the CA1 and area 8b/9 were assessed. The number of branches (slab segments, usually connecting end-points, end-points and junctions or junctions and junctions), the number of actual junctions (merging neighbor junction voxels), and the total length of branches, were measured as main results.

2.4-. Cell loading, data acquisition, and quantitative analysis of spine density and spine morphology

Thick brain sections (200 μm) were used for cell loading, as previously described (Duan et al., 2003). Using DAPI staining (250 ng/ml, Sigma-Aldrich, St. Louis, MO) for 5 min to enable identification of the cortical layers, individual neurons of the pyramidal layer of CA1 and layer 3 of the dlPFC were iontophoretically injected with Alexa Fluor^™ 594 Hydrazide (10 mM in 200 mM potassium chloride, Molecular Probes, Eugene, OR), under a direct current of 3–8 nA until the dye filled the distal ends of the dendrites. Neurons selected for injection were spaced so as to avoid overlapping of their dendrites. Six to ten neurons were injected per section. The sections were mounted with Fluoromount-G (Southern Biotech, Birmingham, AL) on gelatin-coated glass slide. Confocal z- stack images were acquired with an x, y resolution of 0.05 μm and a z step of 0.1 μm using a Plan-Apochromat 63x/1.4 NA oil-immersion objective at 5× zoom, a pinhole setting of 1 Airy unit, and optimal settings for gain and offset, on an upright confocal LSM780 microscope (Carl Zeiss, Jena, Germany). Systematic random sampling was used for the imaging of 6 dendritic segments per neuron (3 basal and 3 apical) in at least 6 neurons per animal. For a dendritic segment to be optically imaged it has to satisfy the following criteria: (1) the entire segment has to fall within a depth of 25 μm; (2) segments have to be either parallel or at acute angles to the coronal surface of the section; and (3) segments do not overlap other segments that would obscure visualization of spines (Duan et al., 2003). After the acquisition, confocal stacks were deconvolved using an iterative blind deconvolution algorithm (AutoQuant X version X3.0.1, MediaCybernetics, Bethesda, MD) and exported to Neurolucida 360 (version 2019.2.1; MBF Bioscience, Williston, VT) to determine spine density and morphology. Neurolucida 360 allows for semiautomated reconstruction of dendrites and spines and classification of spines. Spines with their head diameter (HD): neck diameter (ND) ratio less than 1.4 were classified as stubby. Thin spines had an HD: ND ratio greater than 1.4 and maximum HD less than 0.35 μm. Mushroom spines had an HD: ND ratio diameter greater than 1.4 and maximum HD greater than 0.35 μm. Filopodia had length greater than 2.5 μm.Data from the reconstructions were exported using Neurolucida Explorer (MBF Bioscience, Williston, VT).

2.5-. Statistical analysis

All analyses were performed using GraphPad Prism v.9.3.1, SPSS 21.0 or HLM (v.7.0) and datasets were assessed for normality parameters before significance determination. Experimenters were blinded during imaging experiments. Data are presented in box-and-whiskers plots. Centerlines show the medians, box limits indicate the 25th and 75th percentiles, and whiskers extend 1.5 times the interquartile range (IQR) from the 25th and 75th percentiles. Two-sample unpaired t-test was used for assessing differences between impaired and non-impaired behavioral cohorts. We used the Benjamini, Krieger, and Yekutieli method for correcting for multiple comparisons considering a 5% desired false discovery rate, FDR (Q). The adjusted p-values are reported in the results section. The main effect of each independent variable of the study (behavioral performance and sex of the animal) and its interaction, was also assessed by a two-way ANOVA test, followed by Sidak’s multiple comparisons test. The Kolmogorov-Smirnov test, Bonferroni corrected, was used to compare cumulative frequency distributions, and differences in spine morphometrics were further assessed by Wilcoxon rank sums tests. Spearman correlations were used to study the relationship between aging behavioral trajectories and density and morphological parameters of dendritic spines including age as covariant. Statistical significance was set at an α level of 0.05 in all tests.

3-. Results

3.1-. Behavioral categorization of cognitive aging

Trajectories of cognitive performance over the course of aging were highly variable across individual marmosets (Rothwell, Freire-Cobo, et al., 2021), ranging from continuous improvement or stability to steady decline. The baseline growth models revealed that cognitive change across aging was best described by linear slopes and a quadratic slope did not significantly improve the growth model for SD (χ² (4) = 2.80, p > 0.5) or SR (χ² (4) = 3.96, p > 0.5) test performance across aging. Based on categorization rules, 9 marmosets (5 females, 4 males), were categorized as cognitively impaired (average OLS slope on SD = 15.83, standard deviation [stdev] = 9.09; average OLS slope on SR= 25.83, stdev = 27.13). Whereas, 8 marmosets (4 females, 4 males), were categorized as non-impaired (average OLS slope on SD = −5.27, stdev = 6.64; OLS slope on SR = −19.01, stdev = 9.76). Age at death did not differ between the impaired (average = 8.81 years) and non-impaired (average = 8.37 years) groups (t₍₁₅₎ = 1.132). The slope of change in performance on SD cognitive testing for non-impaired marmosets did not significantly differ from 0 (t₍₁₅₎ = −1.43), indicating no change on average for this group of marmosets across aging. In contrast, the slope of change in performance on SD cognitive testing for impaired marmosets was significantly positive indicating a significant decline in performance across aging (t₍₁₅₎ = 3.11, p = 0.007). The slope of change in performance on SR cognitive testing for non-impaired marmosets was significantly negative (t₍₁₅₎ = −2.54, p = 0.023), indicating that the non-impaired group showed improvement in test performance across aging. The slope of change in performance for the impaired marmosets was marginally positive (t₍₁₅₎ = 2.00, p = 0.064), indicating a decline in performance across aging that did not reach statistical significance. In order to validate these categories, we added cognitive category (impaired, non-impaired) to as a predictor to the linear growth models for trajectories of cognitive performance on SD and SR tests. Initial performance on SD in the first year of testing did not differ between impaired and non-impaired marmosets (intercept: t₍₁₅₎ = −0.89), but cognitive category significantly predicted the change in performance across aging (slope: t₍₁₅₎ = 3.18, p = 0.006) with impaired marmosets showing greater average decline in cognitive performance compared to the non-impaired group (Fig. 2a, supplementary fig. 1a). We compared levels of performance between the two groups at the start of each year of cognitive testing and found that SD performance was similar at years 1, 2, and 3 but significantly worse for the cognitively impaired marmosets at year 4 (t₍₁₅₎ = 2.64, p = 0.018). Likewise, initial performance on SR in the first year of testing was similar across all marmosets regardless of cognitive categorization (intercept: t₍₁₅₎ = −0.493), but the impaired marmosets showed significantly greater decline in cognitive performance across aging compared to non-impaired group (slope: t₍₁₅₎ = 3.22, p = 0.006; Fig. 2b, supplementary fig. 1b). We compared levels of performance between the two groups at the start of each year of cognitive testing and found that SR performance was similar at years 1 and 2 but significantly worse for the cognitively impaired marmosets at years 3 (t₍₁₅₎ = 2.46, p = 0.027) and 4 (t₍₁₅₎ = 3.22, p = 0.006). Age at the start of cognitive testing did not significantly influence initial performance in the first year of testing, nor the slope of performance across years for either SD or SR.

Figure 2. Cognitive behavioral trajectories in aging marmosets.

Plot of performance on SD tests (A) and SR tests (B), measured by trials to criterion, across aging from the beginning of the study (5 years old on average) to the end (9 years old on average). Lines represent the average performance trajectory across all marmosets within each category. Shading represents standard error of the mean. Performance levels were compared across groups at yearly intervals and significant differences are noted. *p < 0.05, **p < 0.01.

3.2-. Neuropathologic findings in the behaviorally impaired marmoset

Mild Aβ burden was observed in both the dlPFC and CA1. More Aβ deposits were found in the cognitively impaired than in non-impaired animals (Fig.3a-d). However this observed difference, did not reach statistical significance in dlPFC (t ₍₁₄₎ = 0.498) or CA1 ( t₍₁₄₎ = 0.594) (Fig. 3e,f). Impaired marmosets also showed a higher astrogliosis particularly noticeable around vascular Aβ deposits (Fig. 4), with higher astrocyte cell densities scored in both dlPFC (t₍₁₄₎ = 0.043), and CA1 (t₍₁₄₎ = 0.042). We did not observe a positive staining of the AT8 antibody that labels hyperphosphorylated forms of the tau protein in any of the marmoset examined. Our negative results might be due to the age of our cohorts (7-9 yo) and that AT8 antibody recorgnizes an advanced hyperphosphorylated state of the tau protein.

Figure 3. Amyloid β burden in the dlPFC and CA1 of aging marmosets.

Confocal image stacks showing: Aβ deposits across dlPFC layers in area 8b/9 of impaired (A) and non-impaired marmoset (B); Aβ deposits in dlPFC, area 8b/9, layer 3 (C) and hippocampal CA1 (D). Different degrees of Aβ burden were observed between impaired animals (top and middle panels in C, top panel in D) and non-Impaired animals (bottom panels in C and D). MOAB-2 (red); DAPI (blue, cyan,white). Relative frequency histograms show distribution of Aβ deposits in the dlPFC (E) and CA1 (F) of impaired , and non-Impaired marmosets. The box-and-whiskers plots showing the differences in Aβ burden (measured as percentage area). Ns, statistically non-significant. Scale bar = 100 μm.

Figure 4. Astrogliosis in the dlPFC and CA1 of cognitively impaired marmosets.

Confocal image stacks show astrocyte proliferation and Aβ deposition in area 8b/9, layer 3 (A) and CA1 (B) of impaired animals. GFAP (green); MOAB-2 (red); DAPI (blue). (C) Quantification of GFAP-expressing astrocytes of the dlPFC and CA1. *, p < 0.05. Scale bars = 100 μm.

We did not find changes in total microglial cell density in dlPFC (t₍₁₄₎ = 0.764) or CA1 (t₍₁₄₎ = 0.594; Fig. 5e). However, there were significant differences on microglial phenotypes between behavioral cohorts and the brain areas analyzed (Fig. 6 and Table 1). Impaired marmosets had a significant increase in the number of intermediate/activated microglia scored in dlPFC compared to non-impaired animals (t₍₁₄₎ = 0.0395), whereas the results between groups did not show statistical significance in CA1 (Table 1). In both brain areas, a higher content of CD68-expressing lysosomes within microglial cells was measured for impaired subjects (Fig. 5d) compared to their non-impaired counterparts in dlPFC (t₍₁₃₎ = 0.042), and CA1 (t₍₁₃₎ = 0.042).

Figure 5. Microglia activation in the cognitively impaired marmoset.

Confocal images of microglial cells in CA1 (A) and layer 3 of dlPFC (B,C). Top panels show merged images of the respective single channels below each column: Iba-1 (green); CD68 (red); and DAPI (blue). In yellow (top panels A-C), CD68-expressing lysosomes within microglial cells. (D) Box-and-whiskers plots showing CD68-expressing lysosome content within Iba-1-immunoreactive microglial cells in the layer 3 of dlPFC and CA1. (E) Quantification of Iba-1 expressing microglia in layer 3 of dlPFC and CA1. *, p < 0.05. Scale bar=10 μm.

Figure 6. Microglial cell morphology in the aging marmoset.

(A) Representation of different microglia cell profiles observed in the marmoset brain: resting (a); Intermediate (b, c), and ameboid and dystrophic (d,e) microglia. Iba-1 (green). Box-and-whiskers plots show the results of the analysis of microglia morphology: process total length (B) and cell branching parameters (C) in layer 3 of dlPFC (left) and CA1 (right). Ns, statistically non-significant; *, p < 0.05. Scale bar=10 μm.

Table 1.

Quantification of microglia cell profiles in layer 3 of dlPFC and CA1

	Impaired	Non-impaired
Resting microglia/mm 2
CA1	37.44 ± 10.59	50.27 ± 16.01
Area 8b/9	31.02 ± 9.78	44.92 ± 14.97
Intermediate microglia/mm 2
CA1	72.06 ± 13.83	57.76 ± 17.14
Area 8b/9	65.25 ± 12.01 *	52.41 ± 6.93
Dystrophic microglia/mm 2
CA1	23.40 ± 6.91	21.39 ± 5.167
Area 8b/9	27.81 ± 8.33	35.56 ± 5.29

Data represent means ± stdev. Significant difference against cognitively non-impaired group (*p < 0.05, t-test).

The results of the morphological analysis of microglia (Fig. 6) showed that microglia in the dlPFC of impaired marmosets had a higher ramified processes with an increased number of branches (t₍₁₃₎ = 0.042), branch endpoints (t₍₁₃₎ = 0.022), and branch junctions (t₍₁₃₎ = 0.022), along with an increased total process length (t₍₁₃₎ = 0.042). No morphological differences between behavioral cohorts were observed in CA1 for any of the same parameters analyzed: branches (t₍₁₃₎ = 0.631); endpoints (t₍₁₃₎ = 0.413); junctions (t₍₁₃₎ =0.413); total process length (t₍₁₃₎ = 0.184).

3.3-. Dendritic spine density and morphology in the behaviorally impaired marmoset

We detected a decrease in total spine density (average number of spines per μm of dendritic length) on basal and apical dendrites in impaired animals (Fig. 7c,d) compared to the non-impaired cohort in the dlPFC (t₍₁₂₎ = 0.048) and CA1 (t₍₁₂₎ = 0.016). This spine loss was specific for the mushroom spines in impaired monkeys in both brain areas analyzed, dlPFC (t₍₁₂₎ = 0.047) and CA1 (t₍₁₂₎ =0.047). Whereas the changes observed in the density of thin spines in dlPFC (t₍₁₂₎ = 0.109) and CA1 (t₍₁₂₎ = 0.109) and stubby spines in dlPFC (t₍₁₂₎ =0.165) and CA1 (t₍₁₂₎ =0.067) did not reach statistical significance. The morphometric analysis of dendritic spines (Fig. 7) showed an increase in the mean head diameter (t₍₁₂₎ =0.0405) and volume (t₍₁₂₎=0.016) of thin spines in dlPFC, with no changes in CA1 (thin spine head diameter t₍₁₂₎ =0.841, and volume t₍₁₂₎ =0.231). There were no significant differences between cohorts in other parameters analyzed: mean head surface of thin spines (t₍₁₂₎ = 0.841 in dlPFC, t₍₁₂₎ =0.828 in CA1); mean head surface of mushroom spines (t₍₁₂₎ = 0.841 in dlPFC, t₍₁₂₎ = 0.267 in CA1); mushroom spine diameter (t₍₁₂₎ = 0.239 in dlPFC, t₍₁₂₎ = 0.297 in CA1) or mushroom spine volume (t₍₁₂₎ = 0.209 in dlPFC, t₍₁₂₎ = 0.297 in CA1). To detect subtle differences in spine morphometrics, we also examined the distribution of spine sizes within each behavioral group (Fig.8a,b). The cumulative distribution frequency of mushroom spine head sizes showed a significant rightward shift (toward larger sizes) in the impaired group in both dlPFC and CA1 (Kolmogorov-Smirnov test, Bonferroni-corrected: mushroom head volume p = 0.049 and diameter p = 0.005 in dlPFC; and mushroom head volume p = 0.0245 in CA1). In addition, the impaired cohort showed a significant rightward shift in the cumulative distribution frequency of thin spine diameter (p = 0.007) and volume (p = 0.055) in dlPFC. The Wilcoxon rank sums test comparing head size parameters confirmed that there were significantly fewer of the smallest thin spines (p < 0.001) in the dlPFC and largger mushroom spines (p < 0.001) in the dlPFC and CA1 of the impaired marmoset.

Figure 7. Dendritic spine density in the dlPFC and CA1 of aging marmosets.

Representative dendritic branches showing the different spine morphology from impaired (top panel) and non-impaired (bottom panel) marmoset in layer 3 of dlPFC (A) and CA1 (B). Quantification of total spine density in layer 3 of dlPFC (C) and CA1 (D). Ns, statistically non-significant; *, p < 0.05. Scale bar= 2 μm.

Figure 8. Dendritic spine morphometry in the dlPFC and CA1 of aging marmosets.

Cumulative frequency plots of individual spines in dlPFC (A) and CA1 (B) show that animals segregate based on cognitive aging trajectories for thin and mushroom spine head size. The rightward shift in the curves for the impaired group indicates that there were significantly more spines with large head dimensions and fewer with smaller head dimensions. Average thin spine head volume (C) and diameter (D) were significantly increased in impaired marmoset in dlPFC.

A Spearman’s test was used to assess the monotonic non-linear relationship between the spine morphometric parameters and behavioral aging trajectories in the marmosets (Fig. 9). The results showed a negative correlation between total spine density in the dlPFC and SR scores (r₍₁₀₎ = −0.600, p = 0.039) and thin spine volume in the dlPFC and SR scores (r₍₁₀₎ = −0.666, p = 0.018).

Figure 9. Cognitive SR scores significantly correlate with total synaptic density and thin spine head volume in the dlPFC.

(A) Scatter plot of SR scores vs. Spine density showing that higher synaptic density is predictive of better behavioral outcome. (B) Scatter plot of SR scores vs. Thin spine head volume, showing that bigger head volumes of thin spines (or fewer smaller thin spines) are predictive of a worse behavioral outcome. Cognitively impaired marmosets (gray dots); cognitively non-impaired marmosets (black dots).

3.4-. Sex-related effects on marmoset neuropathology and dendritic spine morphometry

Analyses incorporating sex as a factor were limited by the number of animals available. The results of the two-way ANOVA tests (Table 2) for all the main outcomes in the study of marmoset brain tissue indicated that the interaction between the two independent variables on the study was not significant (sex of the marmoset and behavioral aging trajectories).

Table 2.

Detailed statistical analysis used in the study.

Figure	Comparison	dlPFC	CA1
Fig.3	Behavior x Sex interaction Behavior vs Aβ deposit Sex vs Aβ deposit	F(1,13)= 3.879, p=0.081 F(1,13)= 0.742, p=0.405 F(1,13)= 0.013, p=0.909	F(1,13)=0.248, p=0.627 F(1,13)= 0.029, p=0.866 F(1,13)= 1.654, p=0.222
Fig.4	Behavior x Sex interaction Behavior vs astrocyte Sex vs astrocyte	F(1,13)= 0.033, p=0.574 F(1,13)= 7.640, p=0.017 F(1,13)= 3.905, p=0.071	F(1,13)= 0.106, p=0.750 F(1,13)= 9.501, p=0.009 F(1,13)= 4.028, p=0.067
Fig.5	Behavior x Sex interaction Behavior vs microglia Sex vs microglia	F(1,13)= 1.456, p=0.250 F(1,13)= 0.025, p=0.875 F(1,13)= 0.078, p=0.784	F(1,13)= 0.300, p=0.593 F(1,13)= 0.319, p=0.582 F(1,13)= 0.853, p=0.373
Fig.5	Behavior x Sex interaction Behavior Vs. CD68/Iba-1 ratio Sex Vs. CD68/Iba-1 ratio	F(1,13)= 0.017, p=0.896 F(1,13)= 5.336, p=0.0414 F(1,13)= 1.505, p=0.2455	F(1,13)= 4.111, p=0.0675 F(1,13)= 8.533, p=0.013 F(1,13)= 4.142, p=0.066
Fig.6	Behavior x Sex interaction Behavior vs branch total length Sex vs branch total length	F(1,11)= 3.576, p=0.0852 F(1,11)= 9.204, p=0.011 F(1,11)= 0.017, p=0.8986	F(1,11)= 0.3905, p=0.544 F(1,11)= 4.755, p=0.0518 F(1,11)= 0.935, p=0.354
Fig.6	Behavior x Sex interaction Behavior vs microglial endpoint Sex vs microglial endpoints	F(1,11)= 3.414, p=0.0917 F(1,11)= 7.839, p=0.017 F(1,11)= 0.018, p=0.893	F(1,11)= 0.432, p=0.524 F(1,11)= 0.06, p=0.805 F(1,11)= 0.018, p=0.895
Fig.6	Behavior x Sex interaction Behavior vs microglial junctions Sex vs microglial junctions	F(1,11)= 2.772, p=0.124 F(1,11)= 5.895, p=0.033 F(1,11)= 0.018, p=0.8986	F(1,11)= 0.274, p=0.611 F(1,11)= 0.248, p=0.628 F(1,11)= 0.208, p=0.656
Fig.6	Behavior x Sex interaction Behavior vs microglial branches Sex Vs. Branches in microglia	F(1,11)= 1.979, p=0.187 F(1,11)= 5.102, p=0.045 F(1,11)= 0.005, p=0.943	F(1,11)= 0.209, p=0.661 F(1,11)= 0.104 p=0.752 F(1,11)= 0.101, p=0.756
Fig.7	Behavior x Sex interaction Behavior vs total spine density Sex vs total spine density	F(1,10)= 0.421, p=0.5308 F(1,10)= 9.866, p=0.0105 F(1,10)= 2.568, p=0.1401	F(1,10)= 0.679, p=0.429 F(1,10)= 9.017, p=0.013 F(1,10)= 0.125 p=0.7308
Fig.7	Behavior x Sex interaction Behavior vs mushroom density Sex vs mushroom density	F(1,10)= 0.097, p=0.761 F(1,10)= 5.679, p=0.038 F(1,10)= 0.833, p=0.382	F(1,10)= 0.025, p=0.876 F(1,10)= 7.533, p=0.0207 F(1,10)= 0.107, p=0.749
Fig.7	Behavior x Sex interaction Behavior vs thin spine density Sex vs thin spine density	F(1,10)= 0.091, p=0.769 F(1,10)= 3.13, p=0.107 F(1,10)= 1.907, p=0.1974	F(1,10)= 0.597, p=0.457 F(1,10)= 4.339, p=0.063 F(1,10)= 0.317, p=0.585
Fig.7	Behavior x Sex interaction Behavior vs stubby spine density Sex vs stubby spine density	F(1,10)= 0.201, p=0.662 F(1,10)= 1.956, p=0.192 F(1,10)= 0.109, p=0.747	F(1,10)= 1.361, p=0.2704 F(1,10)= 0.0027, p=0.959 F(1,10)= 4.781, p=0.067
Fig.8	Behavior x Sex interaction Behavior vs mush head vol Sex vs mush head vol.	F(1,10)= 3.980, p=0.0597 F(1,10)= 4.744, p=0.0508 F(1,10)= 4.213, p=0.631	F(1,10)= 2.630, p=0.136 F(1,10)= 2.063, p=0.1814 F(1,10)= 3.126, p=0.1075
Fig.8	Behavior x Sex interaction Behavior vs thin head vol Sex vs thin head vol	F(1,10)= 0.001, p=0.967 F(1,10)= 8.756, p=0.014 F(1,10)= 2.598, p=0.138	F(1,10)= 0.2809, p=0.607 F(1,10)= 0.8911, p=0.367 F(1,10)= 0.154, p=0.702
Fig.8	Behavior x Sex interaction Behavior vs mush head diameter Sex vs mush head diameter	F(1,10)= 0.741, p=0.409 F(1,10)= 2.743, p=0.128 F(1,10)= 0.459, p=0.513	F(1,10)= 0.9527, p=0.352 F(1,10)= 0.747, p=0.4076 F(1,10)= 1.215, p=0.2962
Fig.8	Behavior x Sex interaction Behavior vs thin head diameter Sex vs thin head diameter	F(1,10)= 0.336, p=0.574 F(1,10)= 5.597, p=0.039 F(1,10)= 0.214, p=0.653	F(1,10)= 0.2261, p=0.644 F(1,10)= 0.0386, p=0.848 F(1,10)= 0.9907, p=0.343
Fig.8	Behavior x Sex interaction Behavior vs mush head surface Sex vs mush head surface	F(1,10)= 0.001, p=0.971 F(1,10)= 0.212, p=0.654 F(1,10)= 0.212, p=0.983	F(1,10)= 0.1350, p=0.720 F(1,10)= 1.815, p=0.207 F(1,10)= 0.0764, p=0.787
Fig.8	Behavior x Sex interaction Behavior vs thin head surface Sex vs thin head surface	F(1,10)= 0.007, p=0.934 F(1,10)= 0.090, p=0.769 F(1,10)= 0.175, p=0.683	F(1,10)= 0.091, p=0.769 F(1,10)= 0.126, p=0.729 F(1,10)= 0.010, p=0.9197

Results from the 2-way ANOVA followed by Sidak’s multiple comparisons test. Statistical significance was set at p≤ 0.05.

Thus, we did not observe an interaction between sex and any of the markers of neuropathology, or the spine morphometrics in any of the behavioral cohorts.

4-. Discussion

We evaluated evidence of neuropathology in the marmoset dlPFC and CA1 and describe structural alterations in dendritic spines during aging of 17 male and female marmosets subjects. We categorized marmosets as cognitively impaired or non-impaired based on their trajectories of cognitive performance over 4 years of aging. Marmosets were categorized as impaired if they showed any sign of cognitive decline, or were otherwise categorized as non-impaired. Aging trajectories significantly differed between impaired and non-impaired groups, suggesting neurodegeneration may occur in some individuals with declining performance.

We found different degrees of Aβ deposition in plaques and vasculature in dlPFC and CA1. Diffuse and compact Aβ₄₂ plaques, have been reported in the neocortex of old (>7 years old) and aged (>15 years old) marmosets, including the sensorimotor, association, and paralimbic regions, but not in the hippocampus, whereas Aβ₄₀ diffuse aggregates were mainly associated with blood vessels (Geula et al., 2002; Rodriguez-Callejas et al., 2016). In the present study, we used the MOAB-2 antibody that detects both Aβ₄₀and Aβ₄₂. We found Aβ deposition in the dlPFC as well as in the CA1 of old animals. Although we did not find statistically significant differences between the behavioral cohorts, the higher degree of Aβ deposition in behaviorally impaired marmosets was concomitant to a significant increased astrogliosis in both brain areas. High numbers of astrocytes have also been reported in old marmosets compared to adolescent and adult individuals (Rodríguez-Callejas et al., 2019). Our results suggest that the modulatory role of glial cells in neuroinflammation, in response to Aβ deposition (Philippens et al., 2017), may contribute to accelerated brain aging process in impaired marmosets.

We did not observe differences in microglia scored between impaired and non-impaired marmosets. However, impaired animals showed a higher CD68-expressing lysosomal content in microglia, in both dlPFC and CA1, which suggests an increased phagocytic activity of these cells. CD68 is a member of the lysosome-associated membrane protein (LAMP) family that is expressed in cells of the monocyte/macrophage lineage (Song et al., 2011). In the brain, CD68 is predominantly expressed in lysosomes of microglia (the brain-resident macrophages) and is used as a marker of activated phagocytic microglia (Ayata et al., 2018; Hendrickx et al., 2017; Mathys et al., 2017). In the dlPFC, we also analyzed different microglial cell phenotypes. A higher number of the intermediate or activated phenotype in detriment of the resting state microglia was observed in impaired marmosets when compared to non-impaired animals. A previous analysis, using a comparable approach, showed different numbers of microglia scored in other regions of the marmoset cortex across the lifespan (Rodríguez-Callejas et al., 2019). Although this study did not include the dlPFC, microglia scores reported for aged marmosets showed reduced values of resting cells, similar to the results for our impaired cohort when compared to the non-impaired animals. This suggests the existence of an aged microglial phenotype in impaired marmosets. Intermediate or activated microglia are described as ramified cells with thick processes and more complex branching, compared to resting cells, characterized by smaller somata and slight ramified branches (Lopes et al., 2008; Rodriguez-Callejas et al., 2016; Streit, 2004; Streit & Xue, 2009). Our morphological analysis confirmed that microglia in impaired marmosets dlPFC had higher process length and increased number of endpoints, junctions, and branches, but not in the CA1. Microglia act as mediator of injury and neurodegeneration, including the clearance of Aβ by phagocytosis (Hughes et al., 2010; Mandrekar et al., 2009; Rogers et al., 2002). Furthermore in aging (Koellhoffer et al., 2017; Streit, 2004; Tay et al., 2017) and neurodegeneration (Krasemann et al., 2017; Salter & Stevens, 2017; Spittau, 2017) these cells may become more reactive or atrophic, explaining regional differences in microglial activation and morphology at different ages and in pathological conditions (Hughes etal.,2010; Koellhoffer et al., 2017; Krasemann et al., 2017; Mandrekar et al., 2009; Rogers et al., 2002; Salter & Stevens, 2017; Spittau, 2017; Streit, 2004; Tay et al., 2017). There was no AT8-labeled tau expression in the brain areas analyzed. Occurrence of tau pathology in the marmoset brain is scarce and rather controversial in the literature. Several key factors need to be considered when interpreting results on presence or absence of hyperphosphorylated tau, including the age of the animals, the method of brain fixation, and the antibodies used for the detection of tau hyperphosphorylated state in marmosets. Our negative results might be in part explained by the age of our animal cohorts (7-9 years). Low levels of abnormally hyperphosphorylated tau (p-T231 tau and p-T212/S214 tau) have been described in the, CA3 field of the hippocampus, the entorhinal and inferior temporal cortex, and parietal cortex, of older marmosets (~11 years), but not in younger adults (~5-6 years) (Rodriguez-Callejas et al., 2016). Conversely, another study did not report abnormally phosphorylated tau (p-S396/S404 tau and p-S262 tau) deposits in very old marmosets (Geula et al., 2002). However, it is important to point out that this study used post-mortem immersion fixation rather than perfusion fixation. Tau dephosphorylates within minutes post-mortem (Sharma et al., 2019) and therefore this method of fixation might not reliably detect phosphorylated tau. A more recent study addressed the issue of the variability in tau isoform detection and phosphorylation in the marmoset brain (Sharma et al., 2019). These authors reported that while marmoset tau conserves the phosphorylation sites seen in humans, it lacks 10 amino acids in the N-terminal motif unique to primates. This might, in part, explain the lack of uniformity when using currently available antibodies for tau detection in NHPs. However, this study used adult marmosets (younger than 6 years of age), and therefore, it remains to be established whether tau phosphorylation patterns change with advancing age in marmosets, resulting in pathological tau aggregation. Here we did not identify tau pathology within the age range our cohort using antibody AT8. Future studies analyzing perfusion-fixed brains of very old marmosets and using an array of different antibodies targeting hyperphosphorylated tau would be needed to assess the extent to which marmosets represent an adequate model to study tau pathology progression during aging.

We also performed a morphometric analysis of dendritic spines in the same brain areas. Our results reveal an age-related loss of dendritic spines on pyramidal cells, with a specific decrease of mushroom spines in impaired marmosets. Interestingly, It has been shown an age-related increase in mushroom spine density in behaviorally unimpaired macaque monkeys compared to impaired animals (Motley et al., 2018), supporting the notion of compensatory mechanism allowing unimpaired monkeys to maintain working memory performance despite thin spine loss in both age groups. In our study, the change in mushroom spine density in this group was also accompanied by a shift in the frequency distribution of mushroom spine sizes in the dlPFC and CA1, with larger spine heads in impaired marmosets. Spine size is directly related to synapse size and larger postsynaptic densities (Harris & Stevens, 1989; Kharazia & Weinberg, 1999; Nusser et al., 1998; Takumi et al., 1999). Larger spines are considered more stable during aging and involved in the storage of long-term memories (Kasai et al., 2003). Our results suggest that the preservation of larger mushroom spines is a compensatory response to the specific decrease in mushroom spine density in impaired animals, in an attempt to maintain cognitive function.

In contrast to mushroom spines, the smaller of the thin spine type are considered less stable and more plastic (Grutzendler et al., 2002; Holtmaat et al., 2005; Kasai et al., 2003; Trachtenberg et al., 2002) to enable rapid dynamic changes in synapses in response to learning (Bourne & Harris, 2007; Kasai et al., 2003). In impaired marmosets we did not find overall changes in thin spine density in the dlPFC or CA1. However, our results are aligned with more in-depth reports of spine morphometrics comparing old versus young macaques (Dumitriu et al., 2010; Luebke et al., 2010; Luebke et al., 2015; Motley et al., 2018; Peters et al., 1994; Peters et al., 2008) as we found larger thin spine head sizes in impaired marmosets in dlPFC. These differences reflect the selective vulnerability of small, thin spines, which are particularly plastic and linked to learning (Bourne & Harris, 2007; Kasai et al., 2003). Thus, the changes in thin spine sizes between impaired and non-impaired marmosets may be due to a difference in spine turnover rates, even though this is not reflected in total thin spine densities between the two cohorts. Future studies incorporating young marmosets are needed to corroborate the role of these changes in spine plasticity as part of an accelerated aging process in impaired monkeys.

We also investigated the potential link between changes in spine morphometrics to functional synaptic plasticity in dlPFC in the context of cognitive aging. We found a direct relationship between the behavioral scores in cognitive function with the change in total densities in dendritic spines and the reduced frequency of the smaller thin spines in dlPFC. The fact that better SR scores (lower scores) are predicted by higher spine densities and preservation of the smallest thin spines is of particular relevance considering that subpopulations of layer 3 pyramidal neurons in dlPFC that form corticocortical pathways are known to be particularly vulnerable in AD (Bussière, Giannakopoulos, et al., 2003; Bussiere, Gold, et al., 2003; Campbell et al., 1991; de Lima et al., 1990; Hof & Morrison, 1990, 2004; Hof et al., 1995; Morrison & Hof, 1997).Therefore, plastic changes in dendritic spines in pyramidal neurons are key for proper cognitive and memory functions in primates (Duan et al., 2003; Hof et al., 1995; Luebke et al., 2010; Luebke et al., 2015; Peters et al., 1994; Peters et al., 2008). Although cognitive impairment emerged earlier in females than in males in this cohort of animals (Rothwell et al., 2022), sex differences were not detected in the analysis of the aging marmoset brain. The current size of the neuropathological dataset might yield low statistical power to detect sex-dependent effects. A larger study including both sexes will be needed to confirm that age-related brain pathology is independent of sex in this species.

Altogether, our results reflect an increased inflammatory state during aging in the marmoset brain, that is more pronounced in cognitively impaired animals.These data are consistent with an accelerated aging process accompanied by neurodegeneration characterizing the cognitively impaired brain. Future longitudinal studies should leverage the short lifespan of the marmoset to follow individuals from early development to advancing age and identify factors conferring neural vulnerability to aging. Such efforts would allow for the discovery of early interventions to preserve cognition in old age.

Supplementary Material

1

Supplementary figure 1. Plots showing average performance level by group (impaired, non-impaired) on each cognitive task: (A) simple discrimination and (B) simple reversal tasks. Error bars represent standard deviation. Across group comparisons were made at yearly intervals and significant differences are noted. *p < 0.05, **p < 0.01.

Highlights.

• Higher Aβ deposit and significant astrogliosis in cognitively impaired marmosets

• Activated microglia with a more complex branching in cognitively impaired marmosets

• Increased phagocytic activity in microglial cells in cognitively impaired marmosets

• Changes in dendritic spine head size and density in cognitively impaired marmosets

• Spine morphometrics correlated with the reversal learning task performance

Funding:

The present study, Neuronal vulnerability to brain aging and neurodegeneration in cognitively impaired marmoset monkeys was and is supported by research grants from by the National Institute of Aging grants R01 AG046266, and F32 AG064925.