Nanoscale organization is changed in native, surface AMPARs by mouse brain region and tauopathy

Abstract

Synaptic AMPA receptors (AMPARs) on neuronal plasma membranes are correlated with learning and memory. Using a unique labeling and super-resolution imaging, we have visualized the nanoscale synaptic and extra-synaptic organization of native surface AMPARs for the first time in mouse brain slices as a function of brain region and tauopathy. We find that the fraction of surface AMPARs organized in synaptic clusters is two-times smaller in the hippocampus compared to the motor and somatosensory cortex. In 6 months old PS19 model of tauopathy, synaptic and extrasynaptic distributions are disrupted in the hippocampus but not in the cortex. Thus, this optimized super-resolution imaging tool allows us to observe synaptic deterioration at the onset of tauopathy before apparent neurodegeneration.

Introduction

Synaptic plasticity plays a crucial role in learning and memory (1, 2) and is primarily expressed by activity-dependent changes in the number, constitution, and conductivity of α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid receptors (AMPARs) in the post-synaptic membrane(3–5). Thus, alteration in surface- (i.e., membrane-) bound AMPARs at excitatory synapses act as indicators of synaptic strength. Changes in AMPAR signaling are associated with multiple neuronal diseases(6–8).

The highly crowded nature of synaptic proteins and the small distance (~20–30 nm) of the synaptic cleft at a chemical synapse (9) requires the use of high-resolution imaging techniques to study the intricate nanoscale organization of AMPARs. A variety of super-resolution fluorescence imaging techniques have been used to show that AMPARs are organized in post-synaptic nanodomains that are correlated with glutamate release sites at the pre-synaptic membrane (10–12), forming a part of a trans-synaptic nano-column (13). AMPARs are also present at extra-synaptic sites as a reserve pool and can transition between synaptic and extra-synaptic sites rapidly through lateral diffusion along with being endo- or exocytosed during synaptic plasticity to regulate their number at the synapse in response to synaptic activity (4, 5, 14, 15).

Most of these studies, however, have been performed on primary neuronal cultures. While being a good model in vitro system, primary cultures lack the physiologic 3D-brain circuitry critical for learning and memory which is affected by numerous neurologic diseases (16, 17). However, visualizing and discerning the AMPAR distribution in thick brain tissue is difficult. The first challenge is selective labeling of native surface AMPARs. Getting access to native surface AMPARs in neurons deep into the tissue is difficult with conventional antibodies (18). To facilitate the labeling of surface AMPARs, previous studies have attached fluorescent proteins or tags on AMPAR, but these protein modification strategies can suffer from over-expression artifacts(14, 19–21). Even in knock-in (KI) mice (22), the presence of a bulky tag can hamper normal AMPAR function and trafficking (23). A promising new approach uses KI mice with a small biotin acceptor peptide (AP) tag on the GluA2 subunit of AMPARs, but this is limited to GluA2 and requires an additional biotinylation step, making the labeling process complicated (24). Another challenge is achieving high-resolution imaging of surface AMPARs in thick tissue. Depth-dependent optical aberrations and increased scattering due to refractive-index mismatch results in poor resolution (25, 26). Recent advances using adaptive optics and in-situ PSF retrieval (INSPR) methods have achieved <10 nm lateral and <30 nm axial localization precision, but their application has been limited to imaging a single color (27–29).

In this work, we use a recently developed probe, called CAM2 bound to a fluorophore, to specifically label native surface AMPARs (GluA2–4 subunits) in mouse brain slices which are kept alive during labeling (18, 30). The small size of CAM2 allows the probe to diffuse rapidly throughout the 200 mm brain slice compared to conventional antibodies (18). Combining this labeling scheme with two-color 3D direct Stochastic Optical Reconstruction Microscopy (dSTORM) (31, 32) that utilizes adaptive optics and INSPR, we observe the nanoscale organization of native, surface AMPARs, by brain region and in the presence of tauopathy, for the first time in mouse brain slices.

We compared the CA1 region of the hippocampus (here called the “hippocampus”) with the motor and somatosensory cortical regions (called the “cortex”) that lie directly above the hippocampus in coronal sections of 6-month-old Thy1-YFP-H mice brains. These two regions have different roles in memory formation and storage despite spatial proximity (33, 34). We find that the fraction of surface AMPARs organized in clusters, which we find are synaptic, is two-times smaller in the hippocampus compared to the cortex. In PS19 mouse model of tauopathy at 6 months of age, when neurodegeneration or brain atrophy has not yet begun (35), we observe rearrangements in the synaptic clustering organization of surface AMPAR along with destabilization of their nanodomains in the hippocampus, but not the cortex. This may underly learning deficits previously observed at this age.

Results

Visualizing synaptic nanoscale architecture of native surface AMPARs in mouse brain slices

CAM2 conjugated to Alexa Fluor 647 was used to fluorescently label surface AMPARs in 200 µm thick live acute brain slices from Thy1-YFP-H mice at 6 months of age (Fig.1a). Using this mouse line allows us to identify a sparse population of single pyramidal neurons and CAM2-labeled AMPARs associated with them. Keeping the brain slices alive during the labeling step ensures that only the surface AMPARs are being labeled (fig. S1). After the incubation with CAM2, acute brain slices were fixed and further cryo-sectioned down to 30 µm thickness before immunostaining with CF568-labeled nanobody against PSD-95, or with primary and secondary antibodies against Homer-1, after detergent-mediated permeabilization. PSD-95 and Homer-1 are located intracellularly in the post-synaptic density (PSD) and help identify the location of excitatory synapses.

Fig. 1. Nanoscale Organization of native surface AMPARs in mouse brain slices.

a. Schematic of the sample preparation. b. Diffraction limited image of a YFP neuron (yellow); AMPAR (cyan) labeled with CAM2-Alexa647 and PSD-95 (magenta) labeled with a nanobody conjugated to CF568. c. Super-resolution dSTORM reconstruction of AMPAR and PSD-95 from b. d. Zoomed in ROIs on individual YFP spines (white boxes) in b. e. Zoomed in ROIs on individual YFP spines (white boxes) in c. f. AMPAR and PSD-95 clusters from the zoomed in ROIs in e. g. AMPAR nanodomains (navy blue) within the clusters in f. h. YFP neuron (yellow) at a depth of ~ 18 µm in the cortex region of a 30 µm thick Thy1-YFP-H mouse brain slice along with dSTORM reconstruction of AMPARs (cyan) and Homer-1 (magenta). i-k. Zoomed in ROIs of individual YFP spines (white boxes) from h. l. Distance between AMPAR-PSD-95 clusters (n=373 cluster pairs from the cortex from three 6-month-old Thy1-YFP-H mice). The peak of the fitted Gaussian function is at 14 nm with standard deviation of 30 nm (R-square=0.95). m. Distance between AMPAR-Homer-1 clusters (n=165 cluster pairs from one 6-month-old Thy1-YFP-H mouse. The peak of the fitted Gaussian function is at 64 nm with a standard deviation of 51 nm (R-square=0.90). n. Paired correlation function of AMPAR and PSD-95 localizations within a cluster pair (n= 373 pairs). Red: experimental data, Blue: simulated randomized distribution. o. AMPAR nanodomain size. The peak of the fitted Gaussian function is at 69 nm with a standard deviation of 22 nm (R-square=0.90). p. Distance of AMPAR localization from the nearest PSD-95 cluster (n=25 YFP neurons from the cortex from three 6-month-old Thy1-YFP-H mice).

For dSTORM imaging of 30 µm cryosections, we used a combination of strategies including refractive index matching, 3-step stage-drift correction, and static aberration correction using a deformable mirror and INSPR (fig. S2a,b and supplementary materials and methods). This allowed us to achieve <10 nm lateral and <30 nm axial localization precisions throughout the 30 µm thick cryosections (fig. S2c-f and fig. S3). For accurate two-color imaging, we measured and corrected chromatic aberrations as a function of imaging depth (fig. S2g).

We visualized native surface AMPARs and PSD-95 in Thy1-YFP-H mice brain slices. In the diffraction-limited images in the cortex (Fig. 1b), we observe surface AMPAR and PSD-95 colocalized with YFP spines (Fig. 1b). With super-resolution dSTORM imaging (Fig. 1c), we observe single molecule blinking events (referred to as localizations) of surface AMPARs and PSD-95; their distributions are clearly resolved (Fig. 1d, e). We used a modified DBSCAN algorithm to identify clusters of PSD-95 and clusters of AMPAR (Fig. 1f), based on their localization density and minimum size (see supplementary materials and methods) (36). The average distance between AMPAR clusters and PSD-95 clusters was found to be 14 nm (Fig. 1l, Table S4). We also separately labeled the PSD protein Homer-1 (Fig.1h-k). The distance between AMPAR clusters from Homer-1 clusters was found to be 64 nm (Fig. 1m, Table S5). These values agree closely with previously reported distances of PSD-95 and Homer-1 from the post-synaptic membrane measured using electron microscopy (37, 38) and dSTORM (39).

The distribution of localizations within an AMPAR-cluster correlates with the corresponding PSD-95 cluster, as shown by comparison against simulated randomized distributions (Fig. 1n, Table S6). Looking within the AMPAR clusters (Fig. 1f), we observed high-density peaks of AMPAR, which were isolated using another modified DBSCAN algorithm (Fig. 1g; see supplementary materials and methods). Previous studies in dissociated neuronal cultures called these peaks “nanodomains” that form a part of the trans-synaptic nanocolumn with the pre-synaptic neurotransmitter release sites (10, 13). We found the size of AMPAR nanodomains to be 69 +/- 22 nm (standard deviation) in our brain slices (Fig. 1o, Table S7), matching closely with the previously reported values (~ 69.6 nm, interquartile range of 54–93 nm)(10) in dissociated neuronal cultures.

We next measured the distance of each localization of an AMPAR around a PSD-95 cluster along a YFP-positive cortical neuron, with the goal of determining synaptic vs extra-synaptic AMPARs (Fig. 1p). Without the clustering analysis of DBSCAN, we limited the analysis to all AMPARs located within 1 µm of a PSD-95 cluster, thereby assigning them to the closest PSD-95 cluster. It should be noted that we detected PSD-95 clusters on about a third of total spines, thus there could be a potential bias towards counting only the synapses with high degree of PSD-95 enrichment to be considered as clusters. The peak of this distribution lies at 20 nm, which corresponds with the previously measured distance between PSD-95 and AMPAR clusters (Fig. 1l). AMPAR density sharply fell at longer distances from PSD-95 clusters as expected, consistent with the previous reports that surface AMPARs are highly enriched in synapses and are sparsely dispersed in the extra-synaptic sites (1, 5, 11). Therefore, the high-density regions, i.e., clusters of AMPARs identified earlier using DBSCAN, can be defined as synaptic clusters.

Differences in native surface AMPAR distribution in hippocampus vs cortex

One advantage of imaging native surface AMPARs in brain slices, as opposed to dissociated cultures, is the ability to study brain-region specific AMPAR distribution and its regulation. We therefore compared the hippocampus to the cortex of Thy1-YFP mice at 6 months of age (Fig. 2a-d). By dSTORM, the relative density of total AMPAR —determined by the number of localizations per field of view [FOV] (including all neurons, regardless of the presence of YFP)—is 1.9x larger in the hippocampus than in the cortex (Fig. 2e and Table S8). This is consistent with previous histoblot (40) and high-resolution mass spectrometry (41) studies. However, using super-resolution and cluster analysis, we can accurately quantify and compare synaptic clustering of AMPARs. We find that the fraction of AMPARs in synaptic clusters is lower in the hippocampus than the cortex by 2.9x (Fig. 2f). We find a similar trend (2.5x lower) when limiting the analysis to YFP-labeled pyramidal neurons (Fig. 2g). We also tested whether CAM2 concentration played any role in the clustering differences observed. We increased the labeling of CAM2-Alexa 647 concentration from 2 µM to 5 µM in acute brain slices, and intracranially injected CAM2-Alexa 647 (2 µL, 50 µM) near the CA1 region of the hippocampus in mouse (30). In both cases, we observed the same lower AMPAR clustering in the hippocampus compared to the cortex (fig. S4a-c, 2.5x lower for 5 µM CAM2-AF647, 4x lower for intracranially injected CAM2-AF647). Notably, the trend in AMPAR clustering was found to be reversed in primary (rat) neuronal cultures prepared from the hippocampus and the cortex (DIV 16) (fig. S4d), where a 1.5x increase was observed in the hippocampal cultures compared to cortical cultures. Hence, the observed difference in AMPAR synaptic clustering between the hippocampus and the cortex in brain slices is likely independent of the concentration of AMPAR labeling and is different from the trend observed in dissociated primary neuronal culture.

Fig. 2: AMPAR distribution in hippocampus vs Cortex.

a. Surface AMPAR distribution (cyan) along a YFP neuron (yellow) in the cortex of a 6-months old Thy1-YFP-H mouse, along with PSD-95 (magenta). (Scale bar: 2 µm) b. Zoomed in ROIs from (A) (Scale bar: 0.5 µm). c. Surface AMPAR distribution (cyan) along a YFP neuron (yellow) in the CA1 region of the hippocampus of a 6-months old Thy1-YFP-H mouse, along with PSD-95 (magenta). (Scale bar: 2 µm) d. Zoomed in ROIs from c (Scale bar: 0.5 µm). e. AMPAR localizations per imaging field of view (FoV) (p=0.042). f. Fraction of AMPARs in the whole FoV found in clusters (p=0.004). g. Fraction of AMPARs along a YFP neuron found in clusters (p=0.0004). h. Cumulative frequency distribution of the distance between each AMPAR localization and its nearest PSD-95 cluster in the cortex. i. Cumulative frequency distribution of the distance between each AMPAR localization and its nearest PSD-95 cluster in the hippocampus. j. Fraction of AMPARs lying within 140 nm of a PSD-95 cluster (p=0.003). k. Number of AMPAR clusters in the whole FoV (p=0.441). l. Number of AMPAR clusters on a YFP neuron (p=0.506). m. AMPAR localizations per cluster in the whole FoV (p=0.263). (N) AMPAR localizations per cluster on a YFP neuron (p=0.226).

For e-g and k-n: n=3 mice, at least 6 FoVs per mouse per brain region.

For h,i: Averaged over 3 mice, at least 3 FoVs per mouse.

For j: n=3 mice, at least 3 FoVs per mouse per brain region.

***:P<0.001, as determined using paired sample t-test;

**: p<0.01, as determined using paired sample t-test;

*: p<0.05, as determined using paired sample t-test;

n.s.: p>0.05 as determined using paired sample t-test.

Independent of AMPAR cluster analysis, the difference in synaptic distribution between the hippocampus and cortex is also suggested by measuring the AMPAR distance distribution around PSD-95 clusters along a YFP-positive neuron. There is a change in the slope of the cumulative distribution of AMPAR localizations at ~100–140 nm in the cortex (Fig. 2h, fig. S5a) and in the hippocampus (Fig. 2i, fig. S5b). We therefore define this region around a PSD-95 cluster as the synaptic region. The fraction of AMPAR within this 140 nm synaptic region is 1.8x lower for hippocampus compared to the cortex (Fig. 2j). We note that the exact definition of the synaptic region (for example, 100 nm vs 140 nm) doesn’t affect the observed trend (fig. S5c). Combining the increased extrasynaptic fraction with increased number of total AMPAR (Fig. 2e) in the hippocampus, this means the number of extra-synaptic AMPAR (= total number*extrasynaptic fraction) is higher in the hippocampus compared to the cortex.

We also measured the number of synaptic clusters over the FOV or along a YFP-positive neuron as well as the number of AMPAR within each cluster. We find no change between the hippocampus and the cortex for either of these measurements (Fig 2k-n). This suggests that the number and distribution of surface AMPAR within synaptic regions are the same for the two regions. It should be noted that clustered AMPARs are synaptic but synaptic does not mean clustered. Therefore, the additional AMPARs in the hippocampus are primarily located in the (sparse) extrasynaptic sites.

Changes in the Native surface AMPAR distribution at the onset of tauopathy

Another important advantage of imaging native surface AMPARs in brain slices is the ability to study their changes and therefore synaptic strength as a function of pathology in adult mice. Here, we focus on tauopathy, which is a leading cause of dementia-related illnesses such as Alzheimer’s Disease (42). We use transgenic PS19 mice that model tauopathy by expressing human tau with P301S mutation associated with the frontotemporal dementia. This mutation induces accelerated pathology including tau hyperphosphorylation, formation of neurofibrillary tangles, synaptic deficits neurodegeneration, and cognitive decline. We crossbred Thy1-YFP-H mice with PS19 mice, such that the resulting PS19 mice (referred to as “PS19”) also expresses YFP in a subset of forebrain excitatory pyramidal neurons. At 6 months of age, there is no apparent loss of neurons compared to Thy1-YFP-only mice (referred to as “WT”), but long-term potentiation of excitatory synaptic strength and cognitive deficits have been reported via electrophysiology and behavioral studies. (35, 43).

To shed more light on the effect of tauopathy on synaptic strength, we observed how surface AMPAR organization changes in these mice. As seen in Fig. 3a and 3b showing WT and PS19 hippocampus, the difference is not very obvious, but subtle and important changes are present. The relative localization density of total AMPAR (as previously described for Fig. 2e) —is 1.6x reduced in the hippocampus of PS19 compared to the WT (Fig. 3c, p = 0.023 and Table S9). This reduction in AMPAR numbers have been observed previously in immunohistological studies (which included both surface as well as internal AMPAR)(35, 40) and immunoblots of membrane fractions (which included surface AMPAR)(40). The relative localization density of total AMPAR in the cortex is also reduced by 1.7x, although the larger variance between mice makes this not statistically significant (Fig. 3c, p=0.084).

Fig. 3: AMPAR organization in PS19 mice.

Surface AMPAR distribution (cyan) along a YFP neuron (yellow) in the hippocampus of a 6-months old a. WT and b. PS19 mouse (Scale bar: 2 µm). c. AMPAR localizations detected per imaging FoV (p=0.084, p=0.023). d. Fraction of AMPARs in the whole FoV found in clusters (p=0.024). e. Fraction of AMPARs on a YFP neuron found in clusters (p=0.035). f. Fraction of AMPARs lying within 140 nm of a PSD-95 cluster (p=0.009). g. AMPAR localizations per cluster over whole FOV (p=0.056). h. AMPAR localizations per cluster along a YFP neuron (p=0.048). i. Ratio of AMPARs in a cluster over the whole FoV found in nanodomains (p=0.001). j. Ratio of AMPARs in a cluster on a YFP neuron found in nanodomains (p=0.002). k. Number of AMPARs per nanodomain over the whole FoV (p=0.051). l. Number of AMPARs per nanodomain in a YFP neuron (p=0.027).

For a-c and e-j: n=3 mice each for WT and PS19, at least 6 FoVs per mouse per brain region.

For d: n=3 mice each for WT and PS19, at least 3 FoVs per mouse.

**: p<0.01, as determined using two-sample t-test (Welch Correction);

*p<0.05, as determined using two-sample t-test (Welch Correction);

n.s.: p>0.05 as determined using two-sample t-test (Welch Correction).

Next, we wanted to observe possible changes in synaptic clustering of AMPARs. However, the decreased number of total AMPAR localizations in the PS19 hippocampus may indicate that the AMPAR clusters are simply undetectable because of their lower density with the DBSCAN parameters used. Indeed, upon visual inspection, we found relatively high-density regions not classified as AMPAR clusters with the DBSCAN parameters used for WT mice (fig. S6). Therefore, we reduced our DBSCAN parameters by 1.6x for the PS19 hippocampus data based on the ratio of average AMPAR localizations per field of view detected in the hippocampus of WT and PS19 mice (for WT, the parameters are unchanged). We find that the fraction of AMPARs in synaptic clusters (either across the whole FOV or along a YFP neuron) is higher in the hippocampus of PS19 than in the hippocampus of WT (Fig. 3d, p=0.02; 3e, p=0.03 and Table S10).

To verify whether the fraction of synaptic AMPAR is indeed increasing in the PS19 hippocampus compared to WT, independent of DBSCAN parameter scaling, we compared the synaptic fraction of AMPARs (within 140 nm of PSD-95 cluster) along a YFP-positive neuron. We observed a significant increase in synaptic fraction (Fig. 3f, p=0.008) consistent with the observation of higher fraction of clustered AMPARs in the PS19 hippocampus (Fig. 3d, e). This suggests that tauopathy is associated with an increase in synaptic fraction and a decrease in extrasynaptic fraction of surface AMPARs in the hippocampus.

Given that the PS19 hippocampus has a lower total of number of AMPAR per field of view compared to the WT hippocampus (Fig. 3c), the question is where is this decrease coming from? This decrease could be happening due to a decrease in the number of synaptic and/or extrasynaptic AMPAR. If both were decreasing by the same amount, the fractional change would be the same. However, since extrasynaptic AMPAR fraction is decreasing while the fraction of AMPARs in synaptic clusters is increasing (Fig. 3d, e), we propose that the decrease in total number of AMPAR in the PS19 hippocampus is primarily due to a loss in extrasynaptic AMPAR. This agrees with previously reported decrease in extrasynaptic AMPARs in the PS19 hippocampus observed at 10 months of age using SDS-digested Freeze Fractured Replica Labeling (SDS-FRL) (40).

As in Fig. 2k,l, we used cluster analysis to compare the number of AMPAR localizations per synaptic cluster for the hippocampi in WT and PS19 mice. There is a statistically significant difference for clusters in the YFP-labeled neurons (Fig. 3h, p=0.048) but not for all the clusters in the FOV (Fig. 3g, p=0.055). This suggests that the number of AMPARs per synaptic cluster on YFP-positive neurons is likely to be decreasing by a factor of 1.4 in the PS19 hippocampus.

To probe the AMPAR organization within a synaptic cluster, we measured the fraction of AMPARs found in nanodomains over the whole FOV and along YFP neurons. We find that the fraction is lower in the PS19 hippocampus compared to the WT hippocampus (Fig. 3i, p=0.001; 3j, p=0.002). In addition, the number of AMPAR localization per nanodomain is significantly lower along a YFP neuron (Fig. 3l, p=0.027) but not in the whole FoV (Fig. 3k, p=0.051). These results together suggest that the AMPARs are not as well-organized in synaptic nanodomains in PS19 HP compared to WT HP.

Conclusion

In this work, we observed the nanoscale organization of native, surface AMPARs for the first time in adult mouse brain slices using a combination of CAM2 labeling and dSTORM imaging strategies. We were able to observe the synaptic clustering of AMPARs colocalized with the PSD proteins as well as synaptic nanodomains. Since CAM2 does not label GluA1 subunit, our labeling does not show GluA1 homomers, which, while a minority of AMPARs, are calcium permeable AMPARs believed to be involved in memory formation (44).

We analyzed AMPAR distributions as a function of the brain region, comparing the hippocampus and the neighboring motor and somatosensory cortex. For WT, we found a smaller fraction of AMPARs in synaptic clusters in the hippocampus compared to the cortex, despite having a higher AMPAR density in the hippocampus. We furthermore showed that the additional AMPARs found in the hippocampus are primarily located in extra-synaptic sites. The differences in AMPAR distribution might be related to different functions of the two brain regions in learning and memory.

In PS19 mice, we found that the number of surface AMPARs is reduced in the hippocampus at 6 months of age. Excitatory synaptic strength is getting weaker, as indicated by a reduction of AMPAR per synaptic cluster, as well as disruption of nanodomain organization. This synaptic weakening may be an early indicator of the eventual loss of spines and neurons in PS19 mice at a later age (35, 45, 46). We also observe a greater decrease of extrasynaptic AMPARs compared to synaptic AMPARs in PS19 mice. Since extrasynaptic AMPARs are recruited during LTP, this may explain, in part, the learning deficits of 6-month-old PS19 mice before neurological loss.

Funding Statement

National Institutes of Health grant RF1AG083625 (H.J.C., P.R.S.)

National Institutes of Health grant R01NS100019 (H.J.C., P.R.S.)

National Institutes of Health grant R35GM119785 (F.H.).

Data and materials availability:

All data needed to evaluate the conclusions in the paper are present in the paper and/or the Supplementary Materials. All data used in this paper will be made freely available to those who request and provide a mechanism for feasible data transfers (such as physical hard disk drives or cloud storage).