Carolina Aguado, Sara Badesso, José Martínez-Hernández,?, Alejandro Martín-Belmonte, Rocío Alfaro-Ruiz,Miriam Fernández, Ana Esther Moreno-Martínez, Mar Cuadrado-Tejedor, Ana García-Osta, Rafael Luján,*

Abstract Plaques of amyloid-β (Aβ) and neurofibrillary tangles are the main pathological characteristics of Alzheimer’s disease (AD).However, some older adult people with AD pathological hallmarks can retain cognitive function.Unraveling the factors that lead to this cognitive resilience to AD offers promising prospects for identifying new therapeutic targets.Our hypothesis focuses on the contribution of resilience to changes in excitatory synapses at the structural and molecular levels, which may underlie healthy cognitive performance in aged AD animals.Utilizing the Morris Water Maze test, we selected resilient (asymptomatic) and cognitively impaired aged Tg2576 mice.While the enzyme-linked immunosorbent assay showed similar levels of Aβ42 in both experimental groups, western blot analysis revealed differences in tau pathology in the pre-synaptic supernatant fraction.To further investigate the density of synapses in the hippocampus of 16–18 month-old Tg2576 mice, we employed stereological and electron microscopic methods.Our findings indicated a decrease in the density of excitatory synapses in the stratum radiatum of the hippocampal CA1 in cognitively impaired Tg2576 mice compared with age-matched resilient Tg2576 and nontransgenic controls.Intriguingly, through quantitative immunoelectron microscopy in the hippocampus of impaired and resilient Tg2576 transgenic AD mice,we uncovered differences in the subcellular localization of glutamate receptors.Specifically, the density of GluA1, GluA2/3, and mGlu5 in spines and dendritic shafts of CA1 pyramidal cells in impaired Tg2576 mice was significantly reduced compared with age-matched resilient Tg2576 and non-transgenic controls.Notably, the density of GluA2/3 in resilient Tg2576 mice was significantly increased in spines but not in dendritic shafts compared with impaired Tg2576 and non-transgenic mice.These subcellular findings strongly support the hypothesis that dendritic spine plasticity and synaptic machinery in the hippocampus play crucial roles in the mechanisms of cognitive resilience in Tg2576 mice.

Key Words: aging; Alzheimer′s disease; cognitive; hippocampus; immunoelectron microscopy; resilience; synapse

Introduction

Alzheimer’s disease (AD) is the most common cause of dementia in aging humans, and is characterized by progressive cognitive decline and structural brain alterations, such as amyloid-β (Aβ) deposition in neuritic plaques and neurofibrillary tangles formed by tau protein aggregates.Synapse loss,gliosis, and neuronal subpopulation reduction are also observed (DeTure and Dickson, 2019).Among these changes, synapse loss strongly correlates with cognitive decline and occurs early in the disease (Selkoe, 2002; Subramanian et al., 2020; Zhang et al., 2023).Both Aβ oligomers and tau oligomers have been shown to induce synaptic dysfunction (Marcatti et al., 2022).However, these classical hallmarks of AD are not sufficient to predict the disease, as some older adult individuals remain cognitively intact despite presenting Aβ plaques and neurofibrillary tangles consistent with late-stage AD (Bowles et al., 2019; Ahangari et al.,2023).Studying the factors that confer resilience to the disease in such individuals is challenging due to limited brain access.Therefore, transgenic mice, like the Tg2576 model, with AD-linked genetic mutations, provide valuable insights into the molecular mechanisms underlying resilience (Pérez-González et al., 2020, 2021).Approximately 22% of these Tg2576 mice exhibit substantial AD pathology but do not develop memory deficits (Pérez-González et al., 2020, 2021).Notably, higher expression of the phospholipase PLA2G4E in the hippocampus of resilient Tg2576 mice compared with impaired mice may underlie resistance to dementia (Pérez-González et al., 2020).Furthermore, PLA2G4E overexpression in the hippocampus of APP/PS1 mice restored memory deficits and increased dendritic spines without affecting amyloid pathology (Pérez-González et al., 2020).

Pathologically, Aβ has been shown to reduce the density of dendritic spines and excitatory synapses (Spires et al., 2005; Ortiz-Sanz et al., 2020).Aβ also affects long-term potentiation (LTP) and long-term depression (LTD) by modulating glutamate receptor-dependent signaling pathways (Shankar et al., 2007; Li et al., 2009).Overexpression of mutant human APP can alter neurotransmitter receptors and ion channels associated with excitatory synapses, such as AMPA receptors, mGlu5 receptors, and GABAB receptors,despite little or no cell death and no neurofibrillary pathology (Martín-Belmonte et al., 2020a, b, 2021, 2022).Therefore, the loss of synapses and the decrease in the density of glutamate receptors induced by Aβ may underlie the behavioral deficits observed in Tg2576 mice (Dong et al., 2007).

Recently, a study using human post-mortem brain tissue highlighted the importance of synaptic resilience in maintaining cognition during aging (King et al., 2023).Considering the significant contribution of synaptic dysfunction and loss in Alzheimer’s disease (AD) pathology, developing innovative strategies to protect synapses from the toxic effects of Aβ and tau may preserve synaptic health and cognitive function.In this study, we utilized the Morris Water Maze (MWM) test to identify resilient aged Tg2576 mice, as previously described (Pérez-González et al., 2020, 2021), and established a comparison with their impaired and wild type littermates.Employing electron microscopic approaches, we performed a thorough analysis of synaptic pathology and molecular composition to examine the synaptic changes associated with resilience to cognitive decline in aged Tg2576 mice.Our findings provide compelling evidence that synapse density and subcellular distribution of glutamate receptors differ between resilient and cognitiveimpaired Tg2576 mice.

Methods

Animals

Transgenic aged female Tg2576 mice (16–18 months old, 25–30 g) were utilized to identify cognitively resilient mice.All Tg2576 mice used in this study were derived from an in-house colony and they were maintained (breeding protocol 061c-19) on the original hybrid C57BL/6J-SJL line (Janvier Lab, Le Genest-Saint-Isle, France; Hsiao et al., 1996).This strain overexpresses hAPP with the familial Swedish AD mutation (K670N/M671L) under the control of the prion promoter.We always use female Tg2576 mice because males show a very aggressive behavior and present signs of fighting, manifested as bite wounds which can result in death.On the contrary, females can be housed together, get along very well, and always show a more homogeneous behavior.Age-matched transgene-negative littermates (Tg2576–/–) were selected as the control group.Female Tg2576 mice can be housed together,and they get along very well and always show a more homogeneous behavior.On the contrary, males show a very aggressive behavior and present signs of fighting, manifested as bite wounds, which can result in death.The mice were housed in groups of 4–6 per cage with free access to food and water,maintaining a temperature-controlled environment on a 12-hour lightdark cycle.All study procedures adhered strictly to current European and Spanish regulations (2010/63/EU; RD52/2013) and the study was approved by the Ethics Committee of the University of Navarra on November 23, 2018(protocols 113-18).All experiments were designed and reported according to the Animal Research: Reporting ofIn VivoExperiments (ARRIVE) guidelines(Percie du Sert et al., 2020).

For immunoelectron microscopy, the animals were anesthetized via intraperitoneal injection of 2% sodium pentobarbital (50 mg/kg, i.p.;Dolethal?; Vetoquinol, Madrid, Spain).Following the abolition of reflex activity, the heart was surgically exposed for perfusion fixation through the ascending aorta.We used 0.9% saline, followed by freshly prepared ice-cold fixative containing 4% paraformaldehyde, 0.05% glutaraldehyde, and 15% (v/v)saturated picric acid in 0.1 M phosphate buffer (PB, pH 7.4).The brains were then removed, immersed in the same fixative for either 2 hours or overnight at 4°C, and thoroughly washed in 0.1 M PB.Coronal 60-μm thick sections were cut using a Vibratome (Leica V1000, Barcelona, Spain).

Morris Water Maze (MWM) Test

The Morris Water Maze (MWM) test was utilized to select cognitive resilient mice as described earlier (Pérez-González et al., 2021; Additional file 1 for detailed methods).

Determination of Aβ42 levels and Tau in different extracts

The fractionation of tissue to analyze amyloid and tau pathology in cytoplasm and synaptosomes followed a previously described protocol (Xiao et al., 2019)with some adaptations (Additional file 1 for detailed methods).

Immunohistochemistry

Immunohistochemical reactions at the electron microscopic level were conducted using the pre-embedding immunogold method, following the procedure described earlier by Martín-Belmonte et al.(2020a).The primary antibodies used were: a mouse monoclonal anti-p-Tau PHF1(Ser396/Ser404); a mouse monoclonal anti-total Tau (clone T46, Sigma-Aldrich, St.Louis, MO, USA, Cat# T9450, RRID: AB_477595); a rabbit pan-GluA1 subunit antibody (aa.841–907 of mouse AMPAR, Frontier Institute Co., Sapporo, Japan, Cat# Rb-Af690); a polyclonal rabbit antibody anti-GluA2/3 (Chemicon, Temecula, CA, USA, Cat# AB1506), a rabbit anti-mGlu5 polyclonal antibody (aa.1144–1171 of mouse mGlu5; Frontier Institute Co.,Cat# Rb-Af300, RRID: AB_2571802).The preparation, purification, and full characterization of the rabbit anti-GluA1 polyclonal antibody have been described previously (Fukaya et al., 2006).The specificity of the rabbit anti-mGlu5 polyclonal antibody has been validated in mGlu5 KO mice (García-Negredo et al., 2014).The secondary antibodies used were as follows: HRPconjugated anti-mouse antibody (Cell Signaling Technology, Cat# 7076, RRID:AB_330924), goat anti-mouse and anti-rabbit IgG coupled to 1.4 nm gold antibodies (1:100; Nanoprobes Inc., Stony Brook, NY, USA).

Estimation of synaptic density

The density of synaptic contacts in a specific brain region’s volume is a crucial parameter for assessing structural and functional changes associated with age or pathological conditions (DeFelipe et al., 1999).To estimate synapse density,the stratum radiatum of the hippocampal CA1 field in wild type, impaired Tg2576, and resilient Tg2576 mice was analyzed using the unbiased physical dissector technique.Each dissector was composed of micrographs from two adjacent ultrathin sections: a reference section and a look-up section immediately above it.Random photographs were captured using a JEOL JEM-1400Flash electron microscope equipped with a digital imaging system (JEOL sCMOS camera).Excitatory synapses were quantified within the neuropil,excluding neuronal and glial somata and blood vessels.Only asymmetrical synapses (i.e., excitatory glutamatergic synapses) present in the reference section but absent in the look-up section were counted using FIJI (https://fiji.sc).Synapses were identified by the presence of a presynaptic active zone and a prominent postsynaptic density, separated by a synaptic cleft, along with synaptic vesicles in the presynaptic terminal.All assessments were performed on coded photographs by an observer blinded to the experimental groups.The density of synaptic contacts (Nv) per 1 mm3was then calculated using the following formula:

where Qd represents the number of synaptic contacts per dissector; Vd represents the volume of the dissector given by number of dissectors × frame area × section thickness.We used 20 different dissectors for each animal,totaling 60 dissectors for each experimental group.

Quantification and analysis of pre-embedding immunogold data

To determine the relative abundance of GluA1, GluA2/3, and mGlu5 immunoreactivity in various compartments of CA1 pyramidal cells in wild type,impaired Tg2576, and resilient Tg2576 mice, we utilized 60-μm-thick coronal slices processed for pre-embedding immunogold immunohistochemistry,following a procedure similar to previous studies (Martín-Belmonte et al.,2020a; Marron Fernandez de Velasco et al., 2023; Additional file 1 for detailed methods).

Controls

To assess the method’s specificity in electron microscopy procedures, we conducted control experiments where the primary antibodies were either omitted or substituted with 5% (v/v) normal serum of the same species as the primary antibody.In both cases, no selective labelling was observed,confirming the reliability of the method.Furthermore, we compared the labelling patterns obtained with the results from Calbindin (Swant, Marly,Switzerland).Interestingly, only the antibodies against GluA1, GluA2/3, and mGlu5 consistently labelled the plasma membrane, indicating their specificity in targeting the desired components.

Data analysis

To ensure objectivity, all immunoelectron microscopy experiments were conducted in a blinded manner before data analysis.Statistical analyses were performed using GraphPad Prism 8 (GraphPad Software, San Diego, CA, USA,www.graphpad.com).When comparing two groups, an unpairedt-test was employed.For analyses involving three groups, a one-way analysis of variance followed by Tukey’s multiple comparisons test or the Newman-Keulspost hoctest was used if the samples fulfilled the normality and variance requirements.If these assumptions were not met, the non-parametric Kruskal-Wallis test followed by Dunn’s multiple comparisonpost hoctest was applied.Moreover,for behavioral tests repeated over time, a two-way analysis of variance with repeated measures was employed.Data were presented as mean ± SEM unless explicitly stated otherwise, and statistical significance was considered whenP< 0.05.

Results

Identification of cognitive resilient Tg2576 mice

Tg2576 mice have been shown to develop senile plaques and experience progressive cognitive deficits, ultimately displaying an established AD phenotype by 16–18 months of age (Westerman et al., 2002).In the hiddenplatform phase of the MWM, Tg2576 mice exhibit significantly worse performance than age-matched wild-type mice (effect of dayF(4,155)= 9.51,P< 0.0001; effect of genotypeF(1,155)= 58.24,P< 0.0001; day × genotype interactionF(4,155)= 50.24,P< 0.0001, two-way ANOVA with repeated measures; Figure 1A).The spatial memory deficit was further confirmed in the retention-phase probe trial conducted on the 5thday (P< 0.001; Figure 1B).However, similar to previous studies (Pérez-González et al., 2020, 2021),we identified a group of Tg2576 resilient mice selected based on their performance in the probe trial (percentage of time spent in the quadrant with the hidden platform) equal or above 35% and their escape latency data on the last day of the invisible platform trial equal or less than 20 seconds.Following these criteria, we distinguished a subgroup of aged Tg2576 mice (n= 5 animals) that not only maintained intact cognitive function in the MWM but also performed significantly better than age-matched wild-type mice(two-way ANOVA with repeated measures;F(4,150)= 56.78,P< 0.0001;F(2,150)=164.6,P< 0.0001;F(4,150)= 9.69,P< 0.0001; Figure 1C).Even in the retention phase, the differences among the groups persisted, with the “resilient”group exhibiting the best cognitive performance (one-way ANOVA with group as a variable;F(2,30)= 94.54,P< 0.0001; Figure 1D).Considering the age of the animals, it is expected that even wild-type animals would have a lower performance in the Morris water maze compared with young animals, and it is possible that the molecular mechanisms exhibited by the resilient mice allow them to overcome the poor performance in the test due to aging.Our subsequent investigation aimed to discern whether resilience is associated with any discernible differences in amyloid and tau pathology at the synaptic level.

Figure 1 ｜Learning and memory abilities of the Tg2576 mouse model in the Morris Water Maze (MWM) test.(A) Spatial learning curves (platform hidden) in the water maze representing the escape latency to the platform for wild-type (WT; n = 9) and Tg2576 (n = 24) mice (two-way analysis of variance with repeated measures, ***P < 0.001, ****P < 0.0001).(B) Probe trial with the platform removed 24 hours after completion of hidden training, confirming the spatial memory deficit characteristic of aged Tg2576 mice (unpaired t-test, ***P <0.001).(C) Escape latency to the hidden platform in the reversal MWM test performed for WT (n = 10) , impaired (Tg2576 Imp; n = 19) and resilient Tg2576 (Tg2576 Res; n =5) mice showing that the latter do not present spatial memory impairment (one-way analysis of variance test followed by Tukey’s multiple comparisons test, ***P < 0.001,WT vs.Tg2576 Imp (impaired Tg2576); $$P < 0.01, $$$P < 0.001, WT vs. Tg2576 Res(resilient Tg2576); ++++P < 0.0001, Tg2576 Imp vs. Tg2576 Res.(D) The percentage oftime spent by Tg2576 Res mice in the correct quadrant during the probe trial on day 5 supports the fact that these mice maintain intact spatial memory retention (one-way analysis of variance test followed by Newman-Kewls post hoc test, ***P < 0.001, WT vs.Tg2576 Imp; $$P < 0.01, WT vs. Tg2576 Res; +++P < 0.001, Tg2576 Imp vs.Tg2576 Res).

Amyloid-β and tau pathology

In our previous study, we found no differences in Aβ and tau pathology that could account for the observed resilience in a previous cohort of Tg2576 mice(Perez-Gonzalez et al., 2020).However, given that Aβ and tau are present in synaptosomes from AD brains (Fein et al., 2008), we embarked on an investigation to determine if variations in the levels of these histopathological markers in the pre- or post-synaptic compartments could potentially play a role in resilience.For this purpose, we isolated cytosolic (S2), presynaptic (S3),and crude PSD (P3) fractions from the hippocampi of mice (n= 3 animals per group) and analyzed Aβ levels and phosphorylated tau (pTau) pathology.

As shown in Figure 2A, no significant differences in Aβ levels were observed between impaired and resilient Tg2576 mice in any of the analyzed extracts (total Aβ), cytosolic (S2), or PSD (P3).Interestingly, however, in the presynaptic (S3) fractions, we detected significantly higher levels of pTau/Tau in the impaired mice compared with the resilient mice group (P< 0.05).These findings suggest that a lower amount of synaptic pTau may be responsible for the improved cognitive performance observed in the resilient Tg2576 mice during the MWM (Figure 2B).Subsequently, we explored whether resilience is associated with differences in the composition of synapses or at the synaptic level.

Figure 2｜ Amyloid-β (Aβ) and Tau pathology analyzed in different cortical (guanidine)or hippocampal brain extracts obtained from resilient and impaired Tg2576 mice sacrificed 24 hours after completion of the probe test.(A) No significant differences in Aβ42 levels measured by the enzyme-linked immunosorbent assay (ELISA) were observed between impaired (Tg2576 Imp) and resilient (Tg2576 Res) transgenic mice in the guanidine cortical extracts (total Aβ42), or in the cytosolic fraction (S2, soluble Aβ42) or in the crude PSD hippocampal fraction (P3).(B)Significant differences in the levels of pTau (PHF1 antibody) and total Tau (T46) measured by western blot analysis were detected in the pre-synaptic (S3) and PSD hippocampal fraction (P3) between resilient (Tg2576 Res) and impaired (Tg2576 Imp) transgenic mice(n = 3 animals per group, *P < 0.05, one-way analysis of variance followed by Tukey’s multiple comparisons test).

Synaptic density in the hippocampus of Tg2576 mice

The quantification of excitatory synapses was performed in the neuropil,excluding neuronal and glial somata, blood vessels, apical dendrites, and myelinated axons.This provided us with a comprehensive understanding of the synaptic organization in the CA1 field and the rate of connectivity.The density of excitatory synapses was determined (Figure 3A–D).Statistical analysis demonstrated a significant difference between wild-type and impaired Tg2576 mice (Kruskal-Wallis test followed by Dunn’s multiple comparisonpost hoc;P< 0.001), as well as between impaired Tg2576 and resilient Tg2576 mice (Kruskal-Wallis test followed by Dunn’s multiple comparisonpost hoc;P< 0.001).However, no significant difference was observed between wild-type and resilient Tg2576 mice (P> 0.05), indicating a 20% decrease in the number of synapses per volume in the impaired Tg2576 mice (Figure 3D).This data reveals important insights into the alterations in synaptic density and connectivity patterns in the context of cognitive impairment and resilience, contributing to a deeper understanding of the synaptic changes associated with Tg2576 mice.

Altered subcellular localization of AMPAR subunits in the hippocampus of Tg2576 mice

In the next step, we utilized pre-embedding immunoelectron microscopic approaches in the CA1 stratum radiatum to investigate whether aged Tg2576 mice exhibit evidence of AD-linked dysregulation of AMPA receptors.These receptors have been known to undergo alterations in response to exposure to Aβ, as previously demonstrated in ultrastructural studies on other animal models of AD (Kamenetz et al., 2003; Martín-Belmonte et al., 2020b; Alfaro-Ruiz et al., 2022).To assess whether Aβ leads to a similar redistribution of GluA1 and GluA2/3 subunits in cognitive impaired Tg2576 and resilient Tg2576 mice, we conducted quantitative pre-embedding immunoelectron microscopy in the hippocampus from the CA1 stratum radiatum.In wildtype mice, GluA1 and GluA2/3 immunoreactivity was detected along the extrasynaptic plasma membrane in spines and shafts of CA1 pyramidal cells,as well as at intracellular sites (Figure 4A and D).In impaired Tg2576 mice,immunoparticles for GluA1 and GluA2/3 were more frequently detected intracellularly (Figure 4B and E).However, in resilient Tg2576 mice, the distribution of the two AMPA receptor subunits increased along the plasma membrane (Figure 4C and F).

To investigate the GluA1 and GluA2/3 densities in dendritic domains of CA1 pyramidal cells (Figure 5), we conducted a quantitative comparison.In impaired Tg2576 mice, the GluA1 density along the plasma membrane in dendritic compartments was significantly reduced (P< 0.01 for spines andP< 0.001 for dendrites, Kruskal-Wallis test; Figure 5A).However, no significant differences were found between wild-type and resilient Tg2576 mice (Figure 5A).These changes in density correlated with the percentage of immunoparticles, which showed similar values between wild-type and resilient Tg2576 mice, but there was a decrease of plasma membraneassociated GluA1 and an increase in immunoparticles associated with intracellular sites in CA1 pyramidal cells (Figure 5B).

In impaired Tg2576 mice, the density of GluA2/3 along the plasma membrane was significantly reduced in both spines and dendrites (P< 0.001 for spines and dendrites, Kruskal-Wallis test; Figure 5C).However, in resilient Tg2576 mice,the density of GluA2/3 was significantly increased compared with wild-type mice (P< 0.05 for spines, Kruskal-Wallis test) and impaired Tg2576 mice (P<0.001 for spines and dendrites, Kruskal-Wallis test; Figure 5C).Immunoparticle percentage changes in plasma membraneversusintracellular sites were also observed in both dendrites and spines (Figure 5D).

Similarly, the density of GluA2/3 along the plasma membrane in spines and dendrites in impaired Tg2576 mice (P< 0.001 for spines and dendrites,Kruskal-Wallis test; Figure 5C).However, GluA2/3 density in resilient Tg2576 mice was significantly increased compared to wild-type mice (P< 0.05 for spines, Kruskal-Wallis test) and impaired Tg2576 mice (P< 0.001 for spines and dendrites, Kruskal-Wallis test; Figure 5C).Immunoparticle percentage changes in plasma membraneversusintracellular sites were also observed both at the level of dendrites and spines (Figure 5D).In summary, the amount of GluA1 in resilient mice was maintained at wild-type levels, while the amount of GluA2/3 was significantly increased.

Altered subcellular localization of mGlu5 receptors in the hippocampus of Tg2576 mice

A previous study has revealed a decrease in the density of mGlu5 receptors in dendritic spines and dendritic shafts of CA1 pyramidal cells in APP/PS1 mice (Martín-Belmonte et al., 2021).To expand on this, our next set of analyses aimed to compare the subcellular localization of mGlu5 receptors associated with excitatory synapses in three groups: wild type, impaired Tg2576, and resilient Tg2576 mice.In wild-type mice, mGlu5 immunoparticles were detected to be present in the extrasynaptic plasma membrane of spines and shafts of CA1 pyramidal cells and were also associated with intracellular sites within the same neuronal compartments (Figure 6A).In impaired Tg2576 mice, we observed a significant reduction in the density of mGlu5 immunoparticles in dendritic compartments (P< 0.001 for spines and dendrites, Kruskal-Wallis test) compared with wild-type mice (Figure 6B and D).This decrease in plasma membrane localization was accompanied by an increase in cytoplasmic localization (Figure 6E).Conversely, in resilient Tg2576 mice, the density of mGlu5 immunoparticles was found to be similar to that of wild-type mice (Dunn’s multiple comparisonpost hoc; Figure 6C and D).Additionally, the percentage of mGlu5 immunoparticles along the plasma membrane and at cytoplasmic sites was also comparable to wild-type mice(Figure 6E).In summary, the amount of mGlu5 receptors in resilient mice was maintained at wild-type levels but was increased compared with cognitive impaired mice.

Figure 5 ｜Compartmentalization of AMPA receptor subunits in CA1 pyramidal cells.(A) Plasma membrane-associated immunogold particle density for GluA1 in spines(Kruskal-Wallis test followed by Dunn’s multiple comparison post hoc test; *P < 0.05,**P < 0.01) and dendrites (Kruskal-Wallis test followed by Dunn’s multiple comparison post hoc; **P < 0.01) of CA1 pyramidal cells from wild-type (WT), Tg2576 Imp and Tg2576 Res mice.(B) Distribution of GluA1 immunoparticles at the plasma membrane and intracellular sites in CA1 pyramidal neuron spines and dendrites from WT,impaired Tg2576 mice (Tg2576 Imp) and resilient Tg2576 mice (Tg2576 Res), expressed as a percentage of total particles (n = 3/group).(C) Plasma membrane-associated immunogold particle density for GluA2/3 in spines (Kruskal-Wallis test followed by Dunn’s multiple comparison post hoc test; *P < 0.05, ***P < 0.001) and dendrites (Kruskal-Wallis test followed by Dunn’s multiple comparison post hoc test; **P < 0.001, ***P < 0.001)from CA1 pyramidal cells from WT, Tg2576 Imp and Tg2576 Res mice.(D) Distribution of immunoparticles for GluA2/3 at the plasma membrane (PM) and intracellular sites (Intra)in CA1 pyramidal neuron spines and dendrites from WT, Tg2576 Imp and Tg2576 Res mice, expressed as a percentage of total particles (n = 3/group).In WT and Tg2576 Imp mice, GluA2/3 was less frequently observed along the extrasynaptic plasma membrane of dendrites and spines, but instead more frequently detected at intracellular sites.

Figure 6 ｜Subcellular localization of the mGlu5 receptors in CA1 pyramidal cells.Electron micrographs of the hippocampal CA1 field showing immunoparticles for mGlu5 in wild type (WT), impaired Tg2576 (Tg2576 Imp) mice and resilient Tg2576 (Tg2576 Res)mice, as detected using a pre-embedding immunogold method.(A–C) In WT and resilient Tg2576 mice, most immunoparticles for mGlu5 were located along the extrasynaptic plasma membrane (arrows) of dendritic shafts (Den) and spines (s) of principal neurons,and less frequently detected at intracellular sites (crossed arrows).In Tg2576 Imp mice,immunoparticles for mGlu5 were mostly located at intracellular sites (crossed arrows),and less frequently located along the plasma membrane (arrows) of dendritic shafts(Den) and spines (s).Scale bars: 500 nm.(D) Plasma membrane-associated immunogold particle density for mGlu5 in spines and dendrites (Kruskal-Wallis test followed by Dunn’s multiple comparison post hoc test; *P < 0.05, **P < 0.01, ***P < 0.001) from CA1 pyramidal cells from WT, Tg2576 IMP and Tg2576 RES mice.(E) Distribution of mGlu5 in CA1 pyramidal neuron spines and dendrites from WT, Tg2576 IMP and Tg2576 RES mice,expressed as a percentage of total particles (n = 3/group) along the plasma membrane and at intracellular (intra) sites.In WT and Tg2576 IMP mice, mGlu5 was less frequently observed along the extrasynaptic plasma membrane of dendrites and spines, but instead more frequently detected at intracellular sites.at: Axon terminal.

Discussion

The urgent need for effective dementia therapies is reshaping the approach to AD research.An intriguing avenue for novel therapies stems from the observation that not all older adults with AD neuropathology suffer cognitive impairment; some maintain relatively normal cognitive function.This phenomenon, known as cognitive resilience (Zolochevska et al., 2018;Willroth et al., 2023), has sparked interest in investigating the unique factors within the brains of these cognitively normal individuals that confer resilience against AD pathology.Considering that dendritic spines, the subcellular compartments where most excitatory synapses form, are proposed to be critical for synaptic plasticity underlying learning and memory (Runge et al., 2020), it is reasonable to assume that these structures play a vital role in cognitive resilience.Therefore, our study aimed to evaluate structural and molecular differences in excitatory synapses of the hippocampus that distinguish impaired Tg2576 mice from resilient Tg2576 mice.Our focus was on the stratum radiatum of the hippocampal CA1 field, where we conducted a detailed characterization of synaptic differences, revealing a reduction in excitatory synapse density established on spines of pyramidal cells in impaired Tg2576 mice compared to resilient Tg2576 and wild-type mice.Additionally,we found reduced density of glutamate receptors such as GluA1, GluA2/3, and mGlu5 in spines and dendrites of pyramidal cells in impaired Tg2576 mice.Notably, the density of GluA2/3 was significantly increased in resilient Tg2576 mice.The maintenance of these functional synapses may be attributed to the reduction of tau pathology observed at the presynaptic level.These findings underscore the importance of molecular changes in excitatory synapses as part of the mechanism of cognitive resilience.Identifying molecules that promote healthy synapses and cognitive function will greatly aid in developing new therapeutic interventions aimed at preventing or delaying the clinical progression of AD.

The hippocampus contains predominantly glutamatergic excitatory synapses,which primarily target the dendritic spines of pyramidal cells (Runge et al.,2020).These spines play a crucial role in memory, learning, and cognition.The spine heads possess a postsynaptic density housing the essential molecular machinery required for postsynaptic neurotransmission.Notably, the loss of these spines appears to be associated with Aβ-induced glutamatergic synaptic depression, suggesting alterations in the excitatory synapses’ glutamate receptors (Kamenetz et al., 2003).Our study focused on the spines in the CA1 field and revealed a significant ～20% reduction in excitatory synapses within the stratum radiatum of impaired Tg2576 mice.To reach this conclusion, we employed conventional electron microscopy and stereological approaches,ensuring accurate and reliable data.These findings align with previous studies conducted on animal models of AD and brain sections from affected patients (Spires et al., 2005; Ortiz-Sanz et al., 2020).The observed decrease in synapse density on spines in impaired Tg2576 mice is consistent with the findings of reduced mRNA expression of multiple synaptic genes (Dickey et al., 2003; Leung et al., 2022).Additionally, recent research has identified the presence of tau at the synapses in AD patients as a significant pathogenic event leading to synapse loss and memory impairment (Colom-Cadena et al.,2023).Hence, the accumulation of tau at the synapses of impaired Tg2576 mice could potentially account for the alterations observed in the structure and/or number of synapses.These results provide valuable insights into the mechanisms underlying synaptic dysfunction in AD and further support the relevance of Tg2576 mice as a model for studying this neurodegenerative condition.

Synapse loss in AD and mouse models of AD has often been assessed using synaptophysin immunostaining.In the case of the Tg2576 AD mouse model,prior analysis using light microscopy indicated no significant difference in synaptophysin immunoreactivity when compared with control animals (Spires et al., 2005).However, another study observed no changes in synaptophysin expression but did find a notable decrease in synaptic density when examined through electron microscopy, which is consistent with our own findings presented here (Dong et al., 2007).The variations in these results can be attributed to the limitations of counting immunoreactive puncta at the light microscopic level, which is less precise compared with the estimation of synapses through electron microscopy.This discrepancy arises because synaptophysin labels a specific protein within the axon terminal, thereby only allowing the counting of synapses with a functional axon terminal, which may not necessarily establish a complete synaptic junction.In contrast, electron microscopy provides a more accurate and comprehensive method to estimate changes in synaptic density.As a result, electron microscopy remains the preferred and most reliable method for accurately assessing alterations in synaptic density.By utilizing this advanced technique, we can gain deeper insights into synaptic changes in AD and its animal models, contributing greatly to the advancement of research and the potential development of therapeutic interventions for this complex neurodegenerative condition.

Previous studies have consistently shown that individuals with AD pathology,but normal cognitive function, exhibit a density of dendritic spines similar to those without AD pathology, suggesting that resilient individuals experience minimal or no synapse loss (Lue et al., 1996; Boros et al., 2017).Our findings align with these reports, as we observed no significant difference in synapse density between resilient Tg2576 mice and non-transgenic control mice.This suggests that resilient Tg2576 mice possess an inherent protective mechanism against the degenerative effects of Aβ pathology, which otherwise lead to the cognitive deficits observed in impaired Tg2576 mice.Despite these promising results, the exact protective mechanism in resilient Tg2576 mice and the molecular pathways involved in dendritic spine loss in impaired mice remain incompletely understood.Among the proteins associated with AD, PSD-95, a major scaffolding protein at the synapse, has been found to be significantly depleted in the brains of AD patients (Gardoni et al., 2009; Shao et al., 2011).However, conflicting findings have been reported regarding the association between PSD-95 levels and synaptic vulnerability to AD.Some studies have suggested that low levels of PSD-95 may be a molecular signature of synaptic vulnerability (Gardoni et al., 2009; Shao et al., 2011).Nevertheless, proteomic and immunoblot studies have not shown significant changes in PSD-95 levels across samples from control, AD, and resilient individuals, raising doubts about this hypothesis (Zolochevska et al., 2018).Interestingly, the latter study identified 15 proteins that constitute a unique proteomic signature of PSDs in resilient individuals, setting them apart from both control subjects and AD patients (Zolochevska et al., 2018).These findings open new avenues for further research to explore the potential involvement of these specific proteins in conferring resilience to synaptic loss and cognitive decline in AD.

In the hippocampus, excitatory synapses are characterized by the presence of AMPA and NMDA ionotropic glutamate receptors, along with metabotropic glutamate (mGlu) receptors, all strategically positioned on dendritic spines(Martín-Belmonte et al., 2020b, 2021; Alfaro-Ruiz et al., 2022).These receptors play crucial roles in two fundamental forms of synaptic plasticity:LTP and LTD of synaptic transmission (Bin Ibrahim et al., 2022).The intricate interplay between LTP and LTD is pivotal for learning and memory processes,and mounting evidence suggests that dysregulation of these mechanisms may directly contribute to the learning and memory deficits observed in individuals with AD (Silva and Martínez, 2023).Within the context of AD research, the Tg2576 mouse model has been invaluable, demonstrating synaptic functional deficits, including impaired LTP in the CA1 field (Jung et al., 2011).Our study builds upon these observations by utilizing quantitative immunogold techniques to show a significant reduction in the density of GluA1 and GluA2/3 subunits of AMPA receptors, as well as the mGlu5 subtype of mGlu receptors, along the neuronal surface of spines of CA1 pyramidal cells in the stratum radiatum, with a concomitant increase at cytoplasmic sites in impaired Tg2576 mice.Our findings are in line with previous studies(Almeida et al., 2005; Tanaka et al., 2019; Sathler et al., 2021; Azarnia Tehran et al., 2022), which have demonstrated that exposure to Aβ results in the internalization of surface GluA1 in dissociated cultured neurons, andin vivostudies using AD mouse models have also shown reduced surface GluA1 levels.Moreover, our earlier work has consistently revealed reductions in AMPA and mGlu5 receptors, as well as NMDA and GABAB receptors, in the hippocampus of other AD-related transgenic mice (Martín-Belmonte et al.,2020a, b, 2021, 2022; Alfaro-Ruiz et al., 2022, 2023).

AMPA receptors, found throughout the brain, are particularly abundant in the hippocampus, where all four subunits are highly expressed (Martín-Belmonte et al., 2020b).Among the hippocampal regions, CA1 pyramidal cells show elevated levels of GluA1–3 subunits (Martín-Belmonte et al.,2020b).The functional properties and trafficking mechanism of AMPA receptors depend on their subunit composition (Diering and Huganir,2018).In the hippocampus, both synaptic and surface AMPA receptors predominantly consist of GluA1/2 and GluA2/3 heterodimers (Wenthold et al., 1996).Notably, calcium-permeable channels are formed by homomeric and heteromeric combinations of GluA1, 3, and 4, while the presence of edited GluA2, when combined with other subunits, gives rise to calciumimpermeable AMPA receptors (Diering and Huganir, 2018).A significant finding in our study is the increased density of GluA2/3 immunoparticles on the surface of CA1 pyramidal cells in resilient Tg2576 mice.While our primary antibody recognized both GluA2 and GluA3 subunits, the observed increment is likely mainly attributed to GluA2 immunoreactivity, given that GluA1/2 heterodimers are most common in the hippocampus (Mansour et al., 2001).GluA2-containing AMPA receptors exhibit enhanced stability at the synapse due to their interaction with synaptic proteins that promote receptor retention in the plasma membrane (Shi et al., 2001).Furthermore,the GluA2 subunit appears to play a role in stabilizing dendritic spines in the hippocampus (Passafaro et al., 2003; Sagliettiet al., 2007).Consequently, our data suggest that the significant increase in GluA2, along with heightened Ca2+impermeability in resilient Tg2576 mice, may contribute to the stabilization of AMPA receptors and excitatory synapses.Supporting the importance of the GluA2 subunit in pathological conditions, prior studies have reported reduced GluA2 expression in a model of tauopathy (Alfaro-Ruiz et al., 2022), and hippocampal CA1 pyramidal neurons containing GluA2 have been shown to be particularly vulnerable following an ischemic insult, resulting in decreased GluA2 expression (Gorter et al., 1997).

This study has some limitations that should be noted.The pre-embedding immunogold method only allows determining the subcellular localization of glutamate receptors along the extrasynaptic plasma membrane of hippocampal CA1 pyramidal cells, but do not provide any information about synaptic localization, which requires the employ of the post-embedding immunogold or the SDS-freeze-fracture labeling technique (Martín-Belmonte et al., 2020b) that could not be applied in the present study.This limitation only applies to GluA1 and GluA2/3, because mGlu5 is not present at synaptic sites (Martín-Belmonte et al., 2021).Although changes in the localization of extrasynaptic AMPA receptors parallel similar changes in synaptic ones in animal models of AD (Martín-Belmonte et al., 2020b) we cannot rule out the possibility that synaptic AMPA receptors are differentially affected.However,this would not affect our conclusion that the levels of glutamate receptors in impaired Tg2576 mice are reduced compared to the resilient mice.Moreover,whether there are changes in synaptic density and the localization of glutamate receptors in interneurons also needs to be further investigated.

In summary, our study highlights distinct features that differentiate impaired Tg2576 mice from resilient Tg2576 mice, including the accumulation of presynaptic tau, a lower density of excitatory synapses, and a reduced density of glutamate receptors along the neuronal surface in the hippocampus.These findings offer valuable insights into the molecular components that may contribute to the preserved cognitive integrity observed in resilient Tg2576 mice.

In the future, strategies aimed at reducing tau accumulation at synapses and/or stabilizing excitatory synapses through their molecular machinery may hold promise for treating AD.Many clinical trials have been conducted using tau-targeted drugs, but it remains unknown which toxic species are primarily responsible for neuronal death and should be the focus of future therapies(Wu et al., 2017).There is evidence of diffuse forms of tau in human AD synapses (Tai et al., 2012; Colom-Cadena et al., 2023), which could be an early event preceding the formation of neurofibrillary tangles (Yoshiyama et al.,2007).Our data specifically suggest that reducing tau at synapses could be a strategy to prevent the decrease in glutamate receptors at excitatory synapses observed in AD conditions, thereby preventing memory decline.By targeting these mechanisms, we can advance our efforts to combat cognitive decline and improve the quality of life for individuals affected by this devastating neurodegenerative disorder.

Acknowledgments:We thank Ms Diane Latawiec for the English revision of the manuscript.We also thank Prof.Peter Davis (Department of Pathology,Albert Einstein College of Medicine, NY, USA) for the donation of the p-Tau PHF-1antibody.Author contributions:All authors had full access to all data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.RL and AGO designed the project; SB performed behavior, western blot and ELISA; CA and JMH performed stereology and immunoelectron microscopy; AMB, RAR and MF performed part of the quantitative analyses;CA, SB, JMH, AMB, RAR, MF, AEMM, MCT, AGO and RL analyzed data; RL and AGO wrote the paper.All authors read and approved the final manuscript.

Conflicts of interest:The authors of this manuscript declare that they have no conflict of interest.

Data availability statement:All data relevant to the study are included in the article or uploaded as Additional files.

Open access statement:This is an open access journal, and articles are distributed under the terms of the Creative Commons AttributionNonCommercial-ShareAlike 4.0 License, which allows others to remix, tweak, and build upon the work non-commercially, as long as appropriate credit is given and the new creations are licensed under the identical terms.

Additional file:

Additional file 1: Detailed methods.