Synaptic Organization of Surface AMPARs Changes by Brain Region and Tauopathy

Reviewer #1 (Public review):

Summary:

This study presents a workflow for nanoscale imaging of endogenous synaptic receptors in intact adult brain tissue. By combining live-tissue chemical labeling of native surface AMPA receptors (AMPARs) using the CAM2 probe, two-color 3D super-resolution microscopy, adaptive optics, in situ point spread function retrieval, and chromatic aberration correction, the authors establish a platform for quantitative super-resolution mapping of synaptic architecture deep within brain tissue.

The authors provide convincing evidence that AMPAR organization differs between the hippocampus and cortex and demonstrate that early tauopathy is associated with selective disruption of extrasynaptic and synaptic nanodomain receptor organization in the hippocampus prior to overt neurodegeneration.

Beyond the specific biological findings, the work represents a significant technical advance that enables quantitative investigation of endogenous synaptic receptor architecture in native brain tissue and will be of broad interest to researchers in synaptic biology, super-resolution microscopy, and neurodegeneration.

Strengths:

(1) Introduces and validates a strategy for nanoscale imaging of endogenous surface AMPARs in adult brain slices using CAM2 labeling and two-color 3D super-resolution imaging.

(2) Achieves 10s of nanometers localization precision in thick tissue and validates synaptic structural measurements against previous literature benchmarks.

(3) Reveals regional differences in AMPAR organization, including a larger extrasynaptic receptor pool in hippocampal regions compared with neighboring cortical regions.

(4) Identifies early, region-specific alterations in AMPAR distribution and nanodomain organization in the PS19 tauopathy model.

Weaknesses:

(1) Statistical power is limited by the relatively small number of animals analyzed, although large numbers of synaptic structures were sampled within each mouse.

(2) Synaptic classification relies on PSD-95 cluster detection criteria that may bias analysis toward more strongly enriched synapses, as acknowledged by the authors.

Reviewer #2 (Public review):

Summary:

This is a technically sophisticated effort to visualize the nanoscale organization of native surface AMPA receptors in adult mouse brain. The authors combine chemical labeling of endogenous AMPARs with 3D, 2-color super-resolution microscopy. They use this to compare hippocampal and cortical AMPAR organization and to characterize changes in a P301S mutant mouse tauopathy model. The authors report a greater density of surface AMPAR localizations in hippocampal CA1 than in two examined cortical regions, as well as a smaller fraction of AMPARs organized in synaptic clusters. They report that PS19 tauopathy is associated with reduced hippocampal AMPAR localization density, a relative depletion of the population classified as extrasynaptic, and reduced organization of receptors within synaptic nanodomains.

Strengths:

The methodological contribution is substantial. The authors developed a valuable set of methods for examining endogenous surface AMPAR organization in brain tissue. The imaging methodology could be useful across many questions in synaptic neuroscience.

Weaknesses:

While the biological observations are interesting, several aspects of the neuroscience interpretation require clarification, and some of the biological conclusions should be weakened unless the measurements can better justify them. I provide several comments intended to strengthen this interesting manuscript.

(1) The authors measure fluorescent localization counts but often interpret these as AMPAR numbers. I believe that individual fluorophores can generate multiple localizations and that labeling efficiency can vary, challenging their frequent interpretation as AMPAR numbers. Importantly, AMPARs are tetrameric complexes with potentially different subunit compositions, so one AMPAR could be observed as 0-4 fluorescent localizations. The uncertainty is particularly important for comparisons of AMPAR abundance per cluster and nanodomain. The authors should consistently distinguish localization density from receptor number and clarify which conclusions depend on assumptions about labeling and the assumptions of photophysical equivalence between experimental conditions. The paper would be more rigorous and the claims more legitimate if the authors were disciplined throughout about using empirical descriptors instead of interpretations.

(2) The distinction between synaptic and extrasynaptic receptors is central to the manuscript. The authors identify synaptic receptors either through density-based clustering or proximity to PSD-95 clusters. However, neither measure directly establishes anatomical synaptic localization. This is particularly consequential because only ~1/3 of morphologically identified YFP-positive spines contain PSD-95 clusters that meet the detection criteria. It seems to me that AMPARs associated with weakly-labeled or undetected synapses could then be erroneously classified as extrasynaptic. PSD-95 detection should be assessed across regions and genotypes, and a sensitivity analysis undertaken to ensure that the conclusions are robust to the choices of clustering and distance thresholds.

(3) The authors suggest that the larger extrasynaptic AMPAR population in the hippocampus provides a reserve for rapid synaptic recruitment and that its reduction in tauopathy constrains plasticity. This is plausible, but receptor abundance does not establish receptor mobility or availability for synaptic capture. The importance of AMPAR surface diffusion for hippocampal LTP and contextual learning has been demonstrated (see Penn et al., Nature, 2017), but the current study does not measure diffusion or receptor exchange. The authors should temper their interpretation and explain the caveats.

(4) The authors infer that there has been a preferential extrasynaptic AMPAR loss from the reduced total AMPAR localization density and the increased synaptic fraction they measured. Because both synaptic and extrasynaptic populations could decrease even when the synaptic fraction increases, they should calculate the synaptic and extrasynaptic localization densities separately for each mouse. This would clarify the magnitude and selectivity of the changes that they report and help to reconcile the potentially contradictory increased synaptic fraction with reduced AMPAR localization density within synaptic clusters and nanodomains.

(5) The authors adjust DBSCAN parameters for PS19 hippocampus in response to reduced overall localization density. Because the resulting clustering measurements are themselves used to establish genotype differences, it is important to demonstrate that the conclusions are robust to parameter selection. Analyses using identical parameter sets across conditions, together with simulations in which WT localization patterns are subsampled to match PS19 localization density, would help determine whether the reported changes reflect biological reorganization rather than density-dependent detection effects. Similar validation is important for the nanodomain analysis. I found that, in general, the authors should have made a greater effort to ensure that their measurements are robust to the parameter choices.

(6) CA1 comparisons with motor/somatosensory cortex require more precise anatomical characterization. What cortical layers and hippocampal dendritic compartments are being compared? These are not homogeneous structures in almost any regard. AMPAR or, more empirically, fluorescence localization density per imaging field is not necessarily equivalent to AMPAR density per dendritic membrane area or per synapse. Differences in neuropil composition, spine density, and synapse density could contribute to the observed differences when there is, in fact, no AMPAR density per dendritic membrane area. Proper anatomical normalization or a stricter comparison of equivalent neuronal compartments would strengthen the interpretation. I believe that the authors are aware of the issue, as they did an analysis along a single dendrite, but this has to be done more thoroughly for them to justify the conclusions they assert.

(7) The main comparisons involve three mice per group, and the authors correctly analyze mouse-level averages. Nonetheless, the small number of biological replicates limits the strength of the conclusions about regional specificity and disease effects. The test statistics and degrees of freedom, not just the p-values, should be presented. In addition, effect sizes and individual mouse values should be presented. The absence of a statistically significant cortical effect does not establish that cortex is unaffected. There could be genotype-by-region interactions, which the authors should check.

(8) The authors describe that 6-mo PS19 mice precede neurodegeneration and they assert that the observed AMPAR-associated changes relate to the onset of synaptic dysfunction. However, Yoshiyama et al., Neuron, 2007 demonstrate hippocampal synapse loss and impaired synaptic function at three months in PS19 mice, before detecting fibrillary tau. The present observations likely occur after synaptic pathology in thePS19 mice. The authors should distinguish structural correlates of established dysfunction from the mechanisms that may have initiated that dysfunction. More importantly, the assertion that neuronal or synaptic anatomy is preserved in the PS19 mice should be supported by appropriate measurements.

(9) The reduction of the AMPAR nanodomain organization is potentially important. However, the authors do not measure presynaptic release sites or trans-synaptic alignment. To be rigorous, the data establish altered postsynaptic receptor organization but do not demonstrate disruption of trans-synaptic nanocolumns or establish the functional consequences for quantal transmission. If the authors do not provide evidence for these claims they should not make them and make it clear these are (reasonable) speculations for which they did not collect the data.

Author response:

Reviewer #1 (Public review):

Summary:

This study presents a workflow for nanoscale imaging of endogenous synaptic receptors in intact adult brain tissue. By combining live-tissue chemical labeling of native surface AMPA receptors (AMPARs) using the CAM2 probe, two-color 3D super-resolution microscopy, adaptive optics, in situ point spread function retrieval, and chromatic aberration correction, the authors establish a platform for quantitative super-resolution mapping of synaptic architecture deep within brain tissue.

The authors provide convincing evidence that AMPAR organization differs between the hippocampus and cortex and demonstrate that early tauopathy is associated with selective disruption of extrasynaptic and synaptic nanodomain receptor organization in the hippocampus prior to overt neurodegeneration.

Beyond the specific biological findings, the work represents a significant technical advance that enables quantitative investigation of endogenous synaptic receptor architecture in native brain tissue and will be of broad interest to researchers in synaptic biology, super-resolution microscopy, and neurodegeneration.

Strengths:

(1) Introduces and validates a strategy for nanoscale imaging of endogenous surface AMPARs in adult brain slices using CAM2 labeling and two-color 3D super-resolution imaging.

(2) Achieves 10s of nanometers localization precision in thick tissue and validates synaptic structural measurements against previous literature benchmarks.

(3) Reveals regional differences in AMPAR organization, including a larger extrasynaptic receptor pool in hippocampal regions compared with neighboring cortical regions.

(4) Identifies early, region-specific alterations in AMPAR distribution and nanodomain organization in the PS19 tauopathy model.

We thank the reviewer for reviewing our manuscript and appreciating both the neurological findings, as well as the broader technological advance.

Weaknesses:

(1) Statistical power is limited by the relatively small number of animals analyzed, although large numbers of synaptic structures were sampled within each mouse.

We agree that despite sampling a large number of synaptic structures and imaging fields per mouse, a sample size of n=3 mice per genotype is a limitation of the study. We will highlight this point in the discussion section.

(2) Synaptic classification relies on PSD-95 cluster detection criteria that may bias analysis toward more strongly enriched synapses, as acknowledged by the authors.

We do agree and acknowledge in the manuscript that PSD-95 cluster detection may bias the synaptic fraction analysis towards more enriched synapses. In addition, we will also compare PSD-95 cluster detection across brain regions and genotypes.

Reviewer #2 (Public review):

Summary:

This is a technically sophisticated effort to visualize the nanoscale organization of native surface AMPA receptors in adult mouse brain. The authors combine chemical labeling of endogenous AMPARs with 3D, 2-color super-resolution microscopy. They use this to compare hippocampal and cortical AMPAR organization and to characterize changes in a P301S mutant mouse tauopathy model. The authors report a greater density of surface AMPAR localizations in hippocampal CA1 than in two examined cortical regions, as well as a smaller fraction of AMPARs organized in synaptic clusters. They report that PS19 tauopathy is associated with reduced hippocampal AMPAR localization density, a relative depletion of the population classified as extrasynaptic, and reduced organization of receptors within synaptic nanodomains.

Strengths:

The methodological contribution is substantial. The authors developed a valuable set of methods for examining endogenous surface AMPAR organization in brain tissue. The imaging methodology could be useful across many questions in synaptic neuroscience.

We thank the reviewer for reviewing our manuscript and appreciating the imaging methodology developed in this manuscript.

Weaknesses:

While the biological observations are interesting, several aspects of the neuroscience interpretation require clarification, and some of the biological conclusions should be weakened unless the measurements can better justify them. I provide several comments intended to strengthen this interesting manuscript.

We thank the reviewer for the in-depth comments which will help strengthen our manuscript. We provide our response to each of them below.

(1) The authors measure fluorescent localization counts but often interpret these as AMPAR numbers. I believe that individual fluorophores can generate multiple localizations and that labeling efficiency can vary, challenging their frequent interpretation as AMPAR numbers. Importantly, AMPARs are tetrameric complexes with potentially different subunit compositions, so one AMPAR could be observed as 0-4 fluorescent localizations. The uncertainty is particularly important for comparisons of AMPAR abundance per cluster and nanodomain. The authors should consistently distinguish localization density from receptor number and clarify which conclusions depend on assumptions about labeling and the assumptions of photophysical equivalence between experimental conditions. The paper would be more rigorous and the claims more legitimate if the authors were disciplined throughout about using empirical descriptors instead of interpretations.

We agree that due to the photo-physical behavior of the fluorophore combined with the tetrameric AMPAR composition and labeling efficiency variations, the AMPAR localizations should not be directly interpreted as absolute AMPAR numbers. We will standardize the wording to “AMPAR localizations” instead of “number of AMPARs / receptors” used at times in the Result and Discussion sections. We will also add an explicit statement of the labeling and photophysical assumptions and of which comparisons depend on photophysical equivalence across conditions. Because all samples are prepared, labeled and imaged identically, relative differences in localization density remain valid. We already note the requirements for absolute AMPAR counting in the Discussion section and will expand that caveat.

(2) The distinction between synaptic and extrasynaptic receptors is central to the manuscript. The authors identify synaptic receptors either through density-based clustering or proximity to PSD-95 clusters. However, neither measure directly establishes anatomical synaptic localization. This is particularly consequential because only ~1/3 of morphologically identified YFP-positive spines contain PSD-95 clusters that meet the detection criteria. It seems to me that AMPARs associated with weakly-labeled or undetected synapses could then be erroneously classified as extrasynaptic. PSD-95 detection should be assessed across regions and genotypes, and a sensitivity analysis undertaken to ensure that the conclusions are robust to the choices of clustering and distance thresholds.

We agree with the concern and acknowledge in the manuscript that PSD-95 cluster detection may bias the synaptic fraction analysis towards more enriched synapses. We also thank the reviewer for additional analysis suggestions. We will include a comparison of PSD-95 cluster detection across regions and genotypes. We note that in fig. S7, we already compare two AMPAR-PSD-95 distance thresholds used to define the synaptic region, showing that the differences in synaptic fraction across regions and genotypes are insensitive to small changes in the threshold choice. In addition, we will perform a broader sensitivity analysis of the proximity-based synaptic fraction measurement across varying distance thresholds and PSD-95 clustering parameters.

(3) The authors suggest that the larger extrasynaptic AMPAR population in the hippocampus provides a reserve for rapid synaptic recruitment and that its reduction in tauopathy constrains plasticity. This is plausible, but receptor abundance does not establish receptor mobility or availability for synaptic capture. The importance of AMPAR surface diffusion for hippocampal LTP and contextual learning has been demonstrated (see Penn et al., Nature, 2017), but the current study does not measure diffusion or receptor exchange. The authors should temper their interpretation and explain the caveats.

We agree. Our data measures spatial distributions and not diffusion or receptor exchange. We will present the reserve-pool interpretation (a pool available for rapid synaptic recruitment) explicitly as a hypothesis motivated by prior work rather than a conclusion, and add the caveat that diffusion or exchange measurements would be needed to test it.

(4) The authors infer that there has been a preferential extrasynaptic AMPAR loss from the reduced total AMPAR localization density and the increased synaptic fraction they measured. Because both synaptic and extrasynaptic populations could decrease even when the synaptic fraction increases, they should calculate the synaptic and extrasynaptic localization densities separately for each mouse. This would clarify the magnitude and selectivity of the changes that they report and help to reconcile the potentially contradictory increased synaptic fraction with reduced AMPAR localization density within synaptic clusters and nanodomains.

We thank the reviewer for this suggestion. We will include a comparison of synaptic and extrasynaptic densities calculated separately for each mouse. This will certainly help clarify the magnitude and selectivity of the reported changes.

(5) The authors adjust DBSCAN parameters for PS19 hippocampus in response to reduced overall localization density. Because the resulting clustering measurements are themselves used to establish genotype differences, it is important to demonstrate that the conclusions are robust to parameter selection. Analyses using identical parameter sets across conditions, together with simulations in which WT localization patterns are subsampled to match PS19 localization density, would help determine whether the reported changes reflect biological reorganization rather than density-dependent detection effects. Similar validation is important for the nanodomain analysis. I found that, in general, the authors should have made a greater effort to ensure that their measurements are robust to the parameter choices.

We will separate the two requested analysis, which serve different purposes.

(i) Identical parameters across conditions: we show in Fig. S9 that identical parameters under-detect AMPAR clusters in the lower density PS19 hippocampus, which is precisely why we scaled the DBSCAN parameters for AMPAR cluster analysis in PS19 hippocampus. Moreover, the proximity-based synaptic fraction measurement, which is independent of any AMPAR cluster analysis, was performed with identical distance thresholds and PSD-95 cluster detection parameters across regions and genotypes. It shows an increase in synaptic fraction in PS19 hippocampus (Fig. 4f).

(ii) Density-matched subsampling: we agree this is a key control. We will randomly down-sample AMPAR localization in wild-type hippocampus to match the localization density with PS19 hippocampus, and then re-run AMPAR cluster analysis. This will help identify PS19 changes that reflect genuine reorganization and those that reflect density-dependent effects.

(6) CA1 comparisons with motor/somatosensory cortex require more precise anatomical characterization. What cortical layers and hippocampal dendritic compartments are being compared? These are not homogeneous structures in almost any regard. AMPAR or, more empirically, fluorescence localization density per imaging field is not necessarily equivalent to AMPAR density per dendritic membrane area or per synapse. Differences in neuropil composition, spine density, and synapse density could contribute to the observed differences when there is, in fact, no AMPAR density per dendritic membrane area. Proper anatomical normalization or a stricter comparison of equivalent neuronal compartments would strengthen the interpretation. I believe that the authors are aware of the issue, as they did an analysis along a single dendrite, but this has to be done more thoroughly for them to justify the conclusions they assert.

We thank the reviewer for highlighting this important concern. We agree that CA1 hippocampus and cortex are not homogeneous structures. We will state the compartments explicitly in the revision. Stratum radiatum layer was imaged in the CA1 region of the hippocampus. In cortex, layer V pyramidal neurons are labeled with YFP in Thy1-YFP-H mice and their apical dendritic field was imaged in layers II/III and IV.

In the paper, we show AMPAR cluster analysis across the whole imaging field of view, as well as analysis restricted to the AMPAR localizations along a YFP dendrite for comparison. In addition, we will also report membrane area-normalized AMPAR localization density as well as the spine density across brain regions and genotypes.

(7) The main comparisons involve three mice per group, and the authors correctly analyze mouse-level averages. Nonetheless, the small number of biological replicates limits the strength of the conclusions about regional specificity and disease effects. The test statistics and degrees of freedom, not just the p-values, should be presented. In addition, effect sizes and individual mouse values should be presented. The absence of a statistically significant cortical effect does not establish that cortex is unaffected. There could be genotype-by-region interactions, which the authors should check.

We agree that despite sampling a large number of synaptic structures and imaging fields per mouse, a sample size of n=3 mice per genotype is a limitation of the study. We will further highlight this point in the discussion section.

We already report all the individual mouse values plotted in Fig. 3-4, in Tables S5, S10 and S11. In addition, we will report test statistics, degrees of freedom and effect sizes. We will also test for genotype-by-region interactions.

(8) The authors describe that 6-mo PS19 mice precede neurodegeneration and they assert that the observed AMPAR-associated changes relate to the onset of synaptic dysfunction. However, Yoshiyama et al., Neuron, 2007 demonstrate hippocampal synapse loss and impaired synaptic function at three months in PS19 mice, before detecting fibrillary tau. The present observations likely occur after synaptic pathology in thePS19 mice. The authors should distinguish structural correlates of established dysfunction from the mechanisms that may have initiated that dysfunction. More importantly, the assertion that neuronal or synaptic anatomy is preserved in the PS19 mice should be supported by appropriate measurements.

Yoshiyama et al., Neuron, 2007 report a decline of synaptophysin and α-synuclein immunoreactivity at 3 months of age, a decline of GluR2/3 immunoreactivity at 6 months of age and reduced basal synaptic transmission and CA1 LTP at 6 months of age in PS19 mice. Whereas the filamentous tau lesions were reported to be developed at 6 months of age and overt neuron loss and brain atrophy was reported by 9-12 months of age. In our manuscript, we already note that functional deficits have been reported by Yoshiyama et al. and others by the 6-months age and do not wish to claim that our reported changes precede synaptic pathology and dysfunction. Our precedence statement is relative to the overt neurodegeneration and brain atrophy that is reported at later stages. We will make this distinction more explicit and update any wording that could be read as framing the AMPAR changes as preceding synaptic pathology and dysfunction. In addition, we will also compare the YFP-labeled spine density across brain regions and genotypes.

(9) The reduction of the AMPAR nanodomain organization is potentially important. However, the authors do not measure presynaptic release sites or trans-synaptic alignment. To be rigorous, the data establish altered postsynaptic receptor organization but do not demonstrate disruption of trans-synaptic nanocolumns or establish the functional consequences for quantal transmission. If the authors do not provide evidence for these claims they should not make them and make it clear these are (reasonable) speculations for which they did not collect the data.

We agree that our data establishes altered postsynaptic receptor organization, but does not directly demonstrate the disruption of trans-synaptic nanocolumn. We will restrict our wording to postsynaptic receptor organization and present any statement about nanocolumns or quantal efficacy explicitly as reasonable speculation for which we did not collect any data.