In this study, we developed SCZ-specific brain assembloids that, we believe, may overcome many limitations in studying the pathogenesis of human SCZ, such as the lack of systems representing different stages of developing human brains with cellular complexity.
I think this is one important crux of the study, and could lead to resources useful for the field.
This plasticity of glial cells is likely to be dictated by SCZ neurons, which accounts for the functional abnormalities and other phenotypic changes found in mix-and-match SCZ forebrain assembloids reconstituted with glial cells derived from healthy individuals.
How much do you think this neuron-driven plasticity is derived from the protocol of adding in glia late in the assembloid process? Could the glial cells be integrating in non-physiological ways because of this? Would you predict that different timing of additional of glia change these results?
all four combinatorial forebrain assembloids containing neurons derived from patients with SCZ (SHH, SHS, SSH and SSS) exhibited functional abnormalities, such as decrease of synaptic density, neural excitability, and synapse transmission (Fig. 3b-d and Supplementary Fig. 6a,b),
Could an alternative explanation here be what I mention below, that astrocytes and microglia are doped into the neuronal system later in "development" and thus are not integrated physiologically into the system? Could you test this by looking at the physiology of glial themselves under different assembloid conditions, such as microglial movement (as you did in the previous preprint?) and astrocyte-specific calcium signaling? A related query would be making sure that what you're counting as calcium spikes are truly only coming from neuronal activity. Fluo-4 is known to preferentially load into cortical astrocytes in acute slices, so some of these calcium bursts may not be strictly neuronal.
Skin was swabbed using a microfiber cosmetic applicator and approximately 40µL of surfactant. Swabbing was performed by rubbing the surfactant-coated applicator tip in a back-and-forth motion for 45 seconds, applying light pressure, over a 2mm-by-1mm rectangular area on lesional psoriasis or eczema skin.
What an interesting study, especially the findings relating to patient biomarkers and predicted therapeutic responsiveness.
A few questions arose for me: 1) for the swabbing procedure, how were the affected areas chosen? Were they on similar body parts/skin types? Did this differ for psoriatic or eczematic skin? This information in the Methods section could be informative for readers. 2) Did you test for intra-patient proteomic consistency by sampling more than once per patient? I'm interested in how intra- and inter-patient heterogeneity might differ from each other, affect clustering, or inform overall conclusions. 3) Similarly, could you compare to swabs/proteomics from non-affected skin areas from the same patients? 4) How do you think disease progression dynamics might also introduce noise in proteomics, i.e. could some of the differences in results be explained by sampling from different timepoints?
microglia-specific connexin-43(which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.
Is there a reason you also wouldn't look at an astrocyte-specific Cx-43 knockout? In brain, this is the major connexin that forms gap junctions across the astrocyte syncytium. I'm also wondering whether it would be informative to do the immunostaining above (Fig 4) for Cx-43 since good antibodies exist.
Boldine inhibited glia activation in the lumbar spinal cord dorsal horn.
Thanks for the interesting study about a naturally derived molecule and potential for addressing an unmet clinical need. For this experiment implicating inhibition of glial activation in after boldine treatment, I think the results would be more robust with the addition of 1) immuno-staining for control glial or other histological markers that would not be expected to change with differential levels of glial activation, and 2) quantification of the fluorescence across conditions for the markers shown here (and those controls) for all animals in each group, to allow the readers to see the variance in the data across animals.
Stability of AstroNet pipeline based on time-lapse experiments.
How do these metrics (and stability) differ between the hippocampus and cortex recordings?
However, to further investigate how the functional astrocytic networks obtained by AstroNetdiffer from the quantification of physical connections made by diffusion through gap junc-tions, we use a pipette filled with biocytin.
How would you predict that pharmacological or genetic manipulation of gap junctions affect the connectivity networks? Have you tested this experimentally?
AstroNet pipeline.
Regarding co-activation on short (second-long) timescales, I'm wondering how you square this with our recent work (https://doi.org/10.1038/s41586-024-07311-5) showing that subcellular activity in astrocytes percolates through the gap junctionally coupled network on the order of minutes, and other astrocyte research showing much longer timescales of activity.
the motor cortex (CTX).
What layers of cortex were you recording from, and how might you expect layers to influence these area-level differences you describe?
brqin
Note typo
From each active astrocyte ROI, we extracted the individual fluorescence time-series.
For this analysis, it seems that all the calcium activity in an individual astrocyte is grouped together to generate the time-series data. If this is the case, how is the subcellular activity that characterizes these astrocytes accounted for, and how might that affect the findings here?
3P-imaging of astrocytic Ca2+-activity in mPFC in vivo
Very cool section! A couple of questions: 1. Would you expect that astrocyte calcium activity in deeper layers in mPFC—either in somata or microdomains—would be similar to deep layers in primary sensory areas, which I believe you cite in your comparisons. If they are indeed similar in the features you describe here, does that give specific hints about functions of astrocyte across cortical areas with different architecture.
We further identified HC-HA/PTX3 as the primary bioactive component responsible for pain inhibition.
This is such an exciting overall result. I'm wondering if you've tested/identified any other bioactive compounds from the same material in addition to HC-HA/PTX3, and/or whether you think there may be other significant contributors to pain inhibition from human birth tissues.
but presented at a much lower level in satellite glial cells (Fig S4A).
Even though CD44 is expressed at lower levels in satellite glia than DRGs (hard to tell magnitude without supplemental figures), do you expect that CD44-dependent mechanisms you describe below in DRGs, like cytoskeletal rearrangement) are also occurring in these glial cells and are contributing to the in vivo effects?
Hind leg-directed scratching of the nape was characterized as a scratch.
It would be helpful to readers to clarify whether individual scratches were quantified by observers or detected automatically, i.e. more experimental details, please!
n=7
Clarify in figure that n = mouse number?
we found that the PSD95-tagRFP and Syn-GFP were closely apposed (Fig. 3G&H), indicating that synaptic connections exist between the RVM and LPBN Tacr1 expressing neurons.
I think some zoomed in images from Fig 3H showing the apposed puncta would really help convince readers that these are indeed closely apposed. In addition, adding a quantifiable metric for distance between the puncta would be very helpful.
intra-plantar AITC suppressed chloroquine-induced itch
Likely outside the scope of the current study, but it would be interesting to see whether any aspect of this described circuit is specifically dependent on chloroquine-induced itch, or whether administration of other pruritogens would have differing outcomes.
As predicted, activation of the PBNGrpr neurons (Nagai et al., 2020) (Fig. 1D) suppressed chloroquine-induced scratching behavior in mice, similar to when PBNTacr1 neurons were stimulated.
Do you have DCZ-only and saline controls in animals with an AAV-mCherry control virus?
We observed the expression of grpr and tacr1 mRNA in the external-lateral part of the LPBN (Fig. 1A top center and right). grpr and tacr1 expressing neurons partially overlapped (Fig. 1A bottom middle).
I'd be interested to see a quantification of this overlap of both populations.
Thus, a molecularly defined intra-brainstem circuit between the LPBN and the RVM plays a pivotal role in pain modulation of itch.
Very exciting finding!
,
Minor point, but I believe this comma should be removed for meaning. Otherwise, the main findings are differently stated than what the data show.
Glutamate signaling in astrocytes causes the release of ATP which stimulates propagation of Ca2+ signals (Hamilton et al., 2008). We propose that a similar pathway regulates contraction behaviours in sponges as described in Fig. 5.
Congrats on the fascinating paper on an organism without a conventional nervous system. I'm interested in the functional analogies you propose to astrocytes (vs neurons) based on the described ATP and glutamate dynamics you describe. Do you think that a comparative approach in terms of other features of nervous system cells—expression profiles, cell morphological features and connectivity (i.e. gap junctions)—would be useful for distinguishing and understanding the function of these sponge cells? Or help us think about how nervous system cells like neurons and astrocytes became more specialized in animals with conventional nervous systems?
the amplitudes of neuronal and astrocyte calcium changes were highly correlated
It looks like other waves features were correlated between neurons and astrocytes beyond amplitude, including duration. Was that indeed the case? May be other features to explore below as you look at correlations across brain regions.
Astrocyte calcium waves from mice of this age exhibited similar properties, consisting of a continuous line of co-active astrocytes that propagated primarily in one direction, visible as diagonals in raster plots of cellular activity (Figure 1D; Movie S2).
This is a great way to visualize the waves. I'm wondering how much gap junctions contribute mechanistically to wave propagation within the developing astrocyte networks.
To determine whether coordinated activation of neurons and astrocytes is conserved in other sensory systems, we used in vivo widefield and two-photon imaging of genetically encoded calcium indicators selectively expressed in astrocytes and neurons to define activity patterns in the SC, a visual processing region of the midbrain that receives direct retinal input.
Congrats on a beautiful, technologically impressive piece of work!
It was notable that 95% of microglial plugs covered <1 μm2 in surface area. However, 3.8% of382plugs were >2 μm2 and appeared to essentially take the place of astrocyte endfeet over large383stretches of the vasculature.
This finding, and the one described in the results showing that a single microglia could form 17 plugs made me wonder whether there was anything specific to the microglial morphology itself that could be used to predict these outliers. Would an unbiased analysis of all 3D reconstructed microglia cluster those that formed particularly large or numerous plugs together? Might these cells have a specific morphology? The potentially transient nature of these plugs/contacts might preclude interesting or interpretable findings, but it could lead you to subgroups of microglia to pay attention to.
coverage of vessels, which if true would likely change the number or size of microglial plugs471observed.59 However, the reviewers of that paper raised some important concerns about472their cryo-fixation method
This question of differences in morphology depending on the type of fixation was lingering as I read the paper, so I'm happy you addressed it directly here. I wonder whether there are any publicly available cryo-fixed datasets (either the original eLife study that you cite or others) where microglia could be ID'ed, to compare with the results in this study.
Microglia of the mouse visual cortex were identified within the publicly available Cortical82MM^3 dataset (https://www.microns-explorer.org/cortical-mm3)
I'm a fan of using publicly available datasets to explore new science. Kudos on this interesting project. One thing I was wondering as I read the paper was how easy it was for you to use and navigate this dataset, since that can be a significant hurdle depending on the data type. Any comment on challenges, or what was made easy, by the organization of this dataset for your purposes could be really useful for others looking to analyze it.
Therapeutic implications
Congrats on this very cool (lol) study and unexpected result! When reading the Therapeutic Implications section of the discussion, I wondered whether any of the side effects of topical RAP application for current patients could be clues to its future therapeutic potential, especially with regards to changes (in either direction) in pain, somatosensation, and pruritus.
The number of scratching bouts towards the injection site during 15-minute periodwas counted.
This is a very exciting study overall, and I'm really interested in the mechanistic interaction between the TRESK channel and MrgprA3+ prurireceptors. For the cheek itch assay, it would be great to know whether scratching bouts were counted manually from videos, or whether you used another technique. Thanks!
In line with this hypothesis, in natural recall there was increased316correlation between neurons and astrocytes in the shocked groups compared to the no-shock group. A317simple linear regression analysis revealed that cross-correlations had a linear relationship to average318freezing in the session. During the artificial recall session, although the Shock-ChR2 group exhibited319increased freezing levels and stereotyped calcium signatures time-locked to freezing characteristic of320natural fear, however, there were no significant differences in maximum cross-correlations. Moreover, we321observed more variability in cross-correlations between animals in both shock groups compared to322natural recall.
This summary is really interesting, and I'd love in the discussion for you to speculate on the cellular and circuit function of correlated activity between these two populations of cells? How does this shed light on how astrocytes and neurons are affecting each other? What extracellular signals do these activity patterns make you think might be driving them?
(C-D) Calcium timeseries for astrocytes (C) and neurons (D)
Minor comment: it would be helpful to the reader to have panel titles for Astrocytes and Neurons in C and D, as well as in subsequent figures.
On the other hand, neuronal calcium characteristics were not significantly different across any148groups (Figure 2L-N).
I may be misunderstanding the panels in 2K–L, but could you explain how the Neutral measure of %dF/F in 2L is around 2, while in K it looks much lower than the shock or the eYFP peaks?
jRGECO1a and GCaMP6f successfully and effectively96expressed in vHPC neurons and astrocytes, respectively (Figure 1B).
Do you think you're getting similar expression levels (% of total cell population) of neurons and astrocytes with the calcium indicators based on histology? Do you think differences here would affect your findings?
In many cases, rather than across domains, placebo effects may transfer between modalities and stimuli within the same domain (e.g., between thermal and mechanical pain, but not from pain to itch 91)
For pain vs itch specifically, it would be interesting to hear how your predictions about cross-modality placebo effects were adjusted based on the results of the current study.
First, participants were introduced to the same cream, once presented as “Prodicaine, an effective pain-relieving drug” and once presented as “a control cream with no effects”.
How do you think the findings of this paper would be similar or different if the placebo were a different modality than a cream, say something injected, or would that not have been possible because of the study design?
Throughout history, placebo effects have been variously considered as mysterious healing forces and tricks played upon the gullible by medical practitioners.
This is a really interesting paper, and as someone outside of the field, I appreciate the clear and compelling background section.
cell-intrinsic
I agree that this interpretation makes sense, but the test with TTX might also mean that non-neuronal cells are playing a role in maintaining oscillatory activity, as they can do in the CNS. The pharmacological inhibition of intracellular calcium you use would also block these widely-described mechanisms in astrocytes, for example.
3minutes gave rise to repeated Ca2+ transients that lasted at least 20 minutes
Did you try application of GRP for shorter than 3 minutes, and if so, did you see less of a persistent response?
and identify GRPR itch neurons as the primary excitatory cell-type that responds tocompound 48/80.
Similar to the comment above, would you expect to see similar results with a pruritogen other than 48/40, in particular one that does not involve histaminergic itch?
We found that the total overlap of GRPR:NK1R:SSTR-dspopulation was significantly larger than predicted if the populations were independent (Figures2E and 2F). Dramatic enrichment was also observed in GRPR:NK1R and GRPR-SSTR-dspopulations
Could you use the amount of activity in each cell type, i.e. a few GCaMP spikes vs ongoing oscillations, to enrich this analysis beyond a binary response vs non-response? For example, are more active cells to one peptide more or less likely to be activated by other peptides?
Figure 2. Diverse itch-causing peptides engage a convergent spinal neuron population
Minor note that "theoretical" is misspelled in panels D and E
Notably, the order of peptide application didnot influence the response profiles that we observed,
What a cool finding and nice control!
>3000 excitatory neurons, 16%(553/3367 neurons
I see the n's in the legend for Fig. 2, but for clarity, I would recommend putting the n's in the manuscript text for mouse and slice number.
Given its widespread nature, it is likely that this neural activity was due to acombination of GRPR neurons responding directly to GRP as well as neurons that are activateddownstream of GRPR neuron activity.
Overall, this paper consists of an elegantly structured set of experiments which are technically challenging and really timely. Thank you for the exciting work! One question here: is it possible here to use a blocker of synaptic transmission (but not spiking) to isolate direct vs indirect responses? Similarly, would widespread activation of neurons via high K+ and sucrose allow you to see what proportion of cells are activatable at all in your prep, to better estimate the denominator of these percentages that you present?
(mouse and cat).
This figure would be helped by labeling the species, and using scale bars to show that they aren't to scale.
Thus, glia as well individual glial cell type (spatial) distributionswere broadly similar in mice and humans, suggesting glia as a whole might share somesimilarities in function across species
Exciting! And a premise many of us that work in mouse models are depending on. I also wonder whether different astrocytic subtypes that have been described as primate-specific may be differentiated in your future assays.
In summary,GNR scale with brain size similarly in APC, prefrontal cortex, and the cerebellum.
As you're analyzing several glial subtypes together here, this finding brings up several questions, including whether these relationships are driven by a particular glial cell type and whether/how different glial cell type number are determined developmentally relative to each other.
Ass
typo alert!
Figure 3: Glia volume densities were constant across species in the APCx.
Another small note for several figures is that the resolution is low and sometimes it is difficult to read labels (here, especially the species names).
(d) Shows the estimates of the total number of glia cell types measured with DAB and Nisslstains. The DAB staining bar is broken into three sections, each section a different color denoting thethree major glia-types: sox9 in grey for astrocytes, iba1 in light blue for microglia, and oligodendrocytes indark blue for oligodendrocytes. The bar on the right denotes the estimate of total number of glia under amm2 of surface measured in Nissl stains. (e) Shows overall glia and each glial cell type distribution acrosslayers. The Nissl bar shows the total number of glia in every layer, and the other three show the individualdistribution of each glia-type in every layer.
A small comment on the figure design here is that using the same colors for type of stain (d) and then layer but not stain (e), is a bit confusing.
First, glia density is constant withinthe same circuit, independent of species or brain size. Second, GNRs increase with brain size atthe same rate in different circuits.
I find the language a little confusing here, especially as it differs between the intro and the results/discussion. Here, I think you're using "circuit" to mean the same thing as brain region, but elsewhere, you use "region". I think consistency would help with clarity.
Previous studies have shown that even thoughastrocytes have large cell bodies
In several mouse brain regions, my experience with patching astrocytes and neurons in ex vivo brain slices (as well as DIC images in ex vivo slices from many other groups) shows that astrocytes have smaller cell bodies than neurons, and in fact, can be identified solely based on their smaller somata. I bet that the Nissl staining here likely reflects real neuron-astrocyte size differences, as observed in unfixed tissue.
Mus musculus (mouse), Rattus novergicus (rat), Cavia porcellus (guineapig), Mustela putoris furo (ferret), Monodelphis domestica (short-tailed opossum), Felis catus(domestic cat), and Homo Sapiens (humans).
It's very cool that you were able to obtain tissue from and analyze so many mammalian species, including humans. If you could speculate, do you think you'd find that the same principles you uncovered also hold for non-mammalian vertebrates?
Therefore, the propagationof activity from distal processes to the central somadominated the average delay maps but occurred inthe presence of other processes that were averagedout in our analysis.
In ex vivo (acute slice) astrocyte calcium imaging work, several groups have reported astrocyte calcium events that propagate both toward and away from the soma. How do you square that with these findings? Do you think that those may be smaller events that are less detectable in vivo, with more light-scattering, etc? How do you think they may be involved in the integration described here?
entripetalintegration
I'm curious to know whether "propagation" and "integration" are interchangeable here, or differently defined.
Therefore, ourobservations support the existing evidence fromother brain areas that increased global astrocyticactivity is triggered by arousal and mediated bynoradrenergic neuromodulation13;
I've interpreted the word "global" to mean "within the recorded astrocytes" in the paper. But here, where you cite a review that includes astrocyte calcium dynamics from other brain regions that differ in some cases, and potentially other definitions of the word "global", it feels slightly misleading.
These findings are consistent with direct.CC-BY-NC 4.0 International licenseavailable under a(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
I think there may be accidental problems with this sentence as it transitions from page 4–5, because I don't understand the meaning.
Interestingly, we found that pupil diameterdisplayed the highest correlation with the globalastrocytic signal,
This is exciting because it's very similar to our finding of high correlation between pupil diameter and astrocyte calcium in cortex (Reitman et al, 2023). I'm also interested in whether you think there are differences in this relationship in hippocampus vs cortex.
ROIs were selected based on mean fluorescence(structural label, top row) and a map of local correlations (functional label, bottom row
I'm interested in how the ROIs don't seem to encompass the outer regions of the astrocyte branches, and thus don't tile the field-of-view. Is that because the fluorescence isn't detectable in these regions?
Recent studies have shown that during wakefulness microglia exhibit reduced motilitycompared to anesthetized states, likely due to suppression by NE38,39,59 . Here we show
One question I had throughout the paper is whether you observed changes in microglia motility in your experiments, both because of any biological significance of motility changes, as well as because of how motility could have affected your calcium dynamics results. I'd love to hear more!
CNO-induced Gi activation in microglia caused astrong reduction of cortical extracellular NE concentration
What a dramatic response of extracellular NE! Do you know if CNO causes this kind of response in animals that don't express the DREADD?
However, CNO-induced Gi activation caused a strong increase in Ca2+ activity,primarily in microglia processes
It's so cool that you saw in vivo calcium changes in the microglia processes with this manipulation. I can see from your analyses that there were significant changes in the processes compared to the soma, but I'm wondering whether you observed changes in the processes somewhat evenly distributed across cells, or whether there were particular microglia that had active processes while other cells did not. I know I might be able to answer this question if I could see the videos, but it could be an interesting descriptive analysis to show in addition to what's already in the figure. In addition, whether there's a change in other features of the calcium signal, like propagation, would also be interesting, and might necessitate an non-ROI-based analysis approach.
mean duration of NREM sleepepisodes
I'm curious whether delta power during NREM sleep changed in addition to duration and % time. Especially since you later show that Gi and Gq activation in microglia have similar effects on NREM sleep, it might be a differentiating feature. For Gi vs Gq activation in astrocytes, we previously observed that Gi affected sleep depth while Gq changed sleep duration.
Note, however, that the effect of these antagonists could be mediated either directly byadrenergic receptors on microglia or indirectly through their effects on neurons and astrocytes,which may in turn affect microglia Ca2+ activity
I think this is an important point to make, and I'm glad you did! You may want to include some citations here, especially because it may not be widely known that astrocytes can express high levels of adrenergic receptors, and that subtypes of astrocytic adrenergic receptors can be activated by dopamine, as well as norepinephrine, in the prefrontal cortex.
we also elevatedmicroglia Ca2+ through Gq-mediated activation of PLC-IP3 signaling
I'd be curious to see how the Gi and the Gq activation differ in their activation of calcium, for example, to be able to compare any summary statistics here to those presented in Fig. 2g–k.
Next, to measure cell type‒specific noradrenergic and calcium signaling in response to tactile stimuli, we crossed our dual-NECa mouse line with mice expressing CaMKIIα-Cre or GFAP-Cre to drive the expression of both GRABNE2m and jRGECO1a in excitatory neurons and astrocytes, respectively; we then performed mesoscopic imaging and measured the change in NE and calcium in response to unilateral whisker stimulation (Figure 5E).
It's very exciting to see the transgenic mouse available, and to compare activity between NE and neurons/astrocytes. A question about crossing to the GFAP-Cre line: Did you confirm that expression was limited to astrocytes, or is it a mixed cell-type population? In others' hands, GFAP-Cre has lead to expression in both astrocytes and neurons, and the inducible Ald1h1-CreER line may be a cleaner option.
mportantly, the functional domains of individual astrocytes identified by ASTRA can then be used302to seed the event-based analysis performed by AQuA 28. This was demonstrated in (Fig. S6), where303we show examples of astrocytic domain identified by ASTRA, which were used as priors to instruct304cell-specific AQuA analysis
This integration of ASTRA to determine morphological features and AQuA to quantify event-based fluorescence features seems particularly useful and exciting. It would be really interesting to hear more about cases you've observed where this does and doesn't work well, and your hypotheses for why; this could help others in understanding the best use-cases for analytical tool integration.
Excitatory neurons exhibited decreased sEPSC activity via cell average and K-S test, while exhibiting no significant difference in sIPSC activities in cell averages and a significant difference via K-S test (Fig. 3e-f). Analysis of inhibitory neurons revealed increased sEPSC amplitudes and decreased sIPSC amplitudes via K-S test, which were not statistically significant when averaged across cells (Fig. 3g-h).
My reading of this is that excitatory neuronal activity in these circuits decreases when astrocytic Gabbr1 is knocked out of astrocytes, but inhibitory activity seems to be less affected. Do you think this points to a homeostatic mechanism by which astrocytes may help regulate excitatory/inhibitory balance in developing cortical circuits? It would be interesting to see if this sort of effect would also occur in adult animals, after circuit maturation. Another interesting follow-up may be to think about how much this sort of homeostatic mechanism is driven by physiology or morphology of the astrocytes.
Gabbr1 regulates Ednrb1 during astrocyte morphogenesis
It's very exciting that you found a mechanism downstream of Gabbr1 that has known cytoskeletal function. Maybe I missed this, but do you have an idea about how activation of Gabbr1 might lead to Ednrb upregulation? (Or why Gabbr1 KO leads to down-regulation?)
we found no changes in spontaneous Ca2+ activity in cortical astrocytes from the Gabbr1-cKO
Did you observe any changes in intracellular calcium besides frequency or delta F? In some of our experiments, we find that most salient changes in calcium are related to the propagation or area of the calcium signal. The negative result in calcium here implies that the spontaneous calcium activity in developing cortical astrocytes does not arise from spontaneous inhibitory (GABAergic) signal, which is interesting in light of previous studies. Do you think there might be a developmental timecourse over which astrocytes become responsive to spontaneous release of GABA?
Together, these observations indicate that input from inhibitory neurons contributes to astrocyte morphogenesis in the developing cortex.
Exciting stuff! In this experiment, since the manipulation is neuronal, it's very cool that you see astrocyte-specific morphological effects. I wonder if there also may be neuronal morphology changes at the subcellular level that could precede or accompany the astrocytic changes.