NogoA-expressing astrocytes limit peripheral macrophage infiltration after ischemic brain injury in primates

Astrocytes play critical roles after brain injury, but their precise function is poorly defined. Utilizing single-nuclei transcriptomics to characterize astrocytes after ischemic stroke in the visual cortex of the marmoset monkey, we observed nearly complete segregation between stroke and control astrocyte clusters. Screening for the top 30 differentially expressed genes that might limit stroke recovery, we discovered that a majority of astrocytes expressed RTN4A/ NogoA, a neurite-outgrowth inhibitory protein previously only associated with oligodendrocytes. NogoA upregulation on reactive astrocytes post-stroke was significant in both the marmoset and human brain, whereas only a marginal change was observed in mice. We determined that NogoA mediated an anti-inflammatory response which likely contributes to limiting the infiltration of peripheral macrophages into the surviving parenchyma.


Reviewers' Comments:
Reviewer #1: Remarks to the Author: Studies in primates are often of crucial value on the way of a project or paradigm toward clinical translation. Here, Boghdadi et al. characterize astrocyte responses in visual cortex gray matter after a local ischemic stroke in marmosets by profiling their transcriptome 7 d. after injury. A number of plasticity and inflammation related genes and gene networks were found to be upregulated, among which the membrane protein Nogo-A (also in human strokes, but much less so in rodents!). Via the known Nogo-A receptor LILRB2/PirB Nogo-A restricts macrophage migration into the peri-infarct parenchyma as shown by knock-out experiments. All these observations have relevance for our understanding of the degenerative and repair processes at sites of CNS tissue destruction e.g. after stroke, as well as for the planned and currently on-going clinical trials with anti-Nogo-A therapeutic agents. They thus address a basic neuroscience audience as well as clinical neuroscientists. The following questions should be considered in revising the manuscript: 1. Where precisely in the visual cortex where the astrocyte nuclei for the transcriptomic analysis sampled: All layers of the gray matter? How close to the stroke core? How restricted to the periinfarct ischemic area/penumbra? 2. Two main populations of reactive astrocytes are described: RTN4/Nogo-A-positive and RTN4negative. Is there any correlation of these two types of cells with their location with regard to the stroke core, the penumbra, the infiltration zone…? 3. Fig. 3 shows a comparison of mouse and marmoset astrocytes. The primate astrocytes appear much larger in size, in addition to being strongly Nogo-A-positive. Is this typical and real or a sampling artifact? 4. A description of how far the upregulation of Nogo-A (and presumably several of the other 600 stroke-regulated astrocyte genes) spreads through the cortex is lacking . Fig 4f suggests a very wide distribution, much wider than the penumbra. What about contralateral areas? Please comment. 5. The observation that also the Nogo-A fragment d20, which binds to receptors other than NgR1 or PirB/LILRB2, inhibits macrophage adhesion/migration is very interesting. The discussion of this should take in to account that very little is known on the detailed composition of the Nogo-A receptor complexes. In analogy to e.g. neurotrophin or Wnt receptors, there is a good chance that different subunits form a multisubunit heteromeric complex. Thus, antibodies against one subunit could easily sterically hinder the binding of other subunits to a different ligand site (e.g. anti-LILRB2 could block binding of d20 to S1PR2 or syndecans). Integrins have only been shown to be indirectly involved. The scheme shown in Fig.5d is misleading. 6. Macrophages have multiple functions in CNS lesions, 'good' and 'bad' ones. Effects of restricting their access to the CNS will depend on the time point after injury, the subpopulation of macrophages etc. 7. Minor point: line 466: Several antibodies against Nogo-A as well as receptor bodies and shRNA and KOs have been used in showing enhanced sprouting and regeneration after CNS lesions (not just IN1).
Reviewer #2: Remarks to the Author: This manuscript from Boghdadi and colleagues investigates the heterogeneity of astrocyte responses in a marmoset model of ischemic stroke, as well as highlighting some mechanistic studies around the putative function of Nogo-A in mediating this response by blocking entrance of peripheral macrophages into the brain following insult. This work represents, to my knowledge, the first such analysis of cell-type specific transcritpomic changes in the marmoset model. Indeed the hypothesis that 'immediate post-stroke blockade of NogoA action may exacerbate brain inflammation' is intriguing, and adds a new layer of complexity to ongoing therapeutic trials using anti-Nogo therapies in spinal cord injury -these new data add a great deal to the literature surrounding Nogo as a purely inhibitory molecule. The authors should also be commended for taking some time to compare key findings across marmoset, rodent, and human -particularly important given their affirmation that the marmoset model provides a more appropriate model of the human ischemic response.
I do however have a few concerns with the data as presented in the manuscript, and would like clarifications or additional details to determine how they would alter the conclusions drawn from the study: 1. single nuclei RNASeq analysis -the assertion that astrocytes from stroke and control brains are distinctly different at the transcriptomic level is big surprise and may suggest the data was not anchored or integrated properly. Where there absolutely no physiological (ie. normal/control) astrocytes left following ischemic stroke? Even though V1 visual cortex was dissected out for downstream nuclei sequencing, it still seems unusual that a such a large region was collected that there are no baseline/normal astrocytes left. There needs to be more comprehensive descriptions of what analysis packages/pipelines were used to generate these final data -for instance, it seems likely that the different datasets (control/ischemia) have not been anchored properly (as stated above) -leading to an artifact of complete dataset separation. Canonical Correlation Analysis (or similar) should be used to anchor the experimental groups based on multiple genes that are shown not to change in individual cell types (e.g. for astrocytes perhaps Aldh1l1, Sox9, etc.). It would also be important to include additional QC data pertaining to gene, UMI, etc counts, and highlight how individual animal samples cluster across these small datasets -it is becoming apparent in other published under-powered datasets that the often individual samples are driving much of the reported heterogeneity and the authors would benefit from showing this is not the case with their own datasets. It is also unacceptable that raw data is not deposited at a public repository such as the NIH GEO.
2. additional to this, the samples sizes are quite low, however given the difficulty of obtaining NHP samples this should be considered in that context. The authors should discuss this limitation. Indeed, the astrocyte capture in both control and ischemic animals is among the lowest of all cell types sequenced (~5%), while the capture rate for non-astrocyte cells, particularly of oligodendrocytes that express a lot of RTN4 (the gene that encodes Nogo-A), is much higher. It was a shame not to see these data also integrated into the manuscript -as they are much better powered for making some of the strong conclusions being drawn here. acutely purified astrocytes -why was this only completed in human and also not in mouse/marmoset? Given the statement that astrocytes were purified and cultured from all three species, the lack of this response data was a shame. This would be important data to validate the primate/human-specific nature of this response. 5. TMEM119 and microglia (line 255) -TMEM119 levels have been reported to drop in microglia in the context of infection, injury, and disease. Can the authors comment on this as an alternate interpretation of the TMEM119-IBA1+ cells they labeled here. Given the authors have nucSeq data on a number of microglia (up to 26% of all collected nuclei in the stroke setting) they could mine these data to see if there are any Iba1+ cells in this population that do not express TMEM119. 6. the authors state in several places (e.g. lines 291, 479, among others) that RTN4/Nogo-A has an effect on 'BBB integrity' however this is not measured anywhere in this manuscript. minor points: 1. at several instances throughout the manuscript the authors state that their sequencing results represent 'primate specific responses'. Can the authors show some meta-analysis that confirms this statement? 4. Fig 3 -a low-mag image of the lesioned area would help to validate this regional-heterogeneity of RTN4+ astrocytes 5. Fig 5, panels j -the Nogo-d20 treated cells have a much larger morphology -is this representative of all treatment repeats, or just this image? The side expansion (including of the nuclei) suggest a profound response to this ligand Reviewer #3: Remarks to the Author: This manuscript reports a study examining astrocyte responses to an ischemic stroke in the cerebral cortex of marmosets, a new world species of non-human primates. The main observations are: (a) Single nucleus RNA-sequencing (snRNAseq) is used to identify and compare 2107 nuclei in stroke tissue from n=3 animals with 594 nuclei from n=3 uninjured controls. From a technical perspective, the procedures are well described and appear rigorously conducted. The data look to be of good quality. The cluster-analyses look robust and the differentially express gene (DEG) data look reliable. (b) From the snRNAseq data the authors noted RTN4A among the top 30 DEGs in stroke associated astrocyte nuclei. RTN4A is a molecule that broadly repulses migrating cells (first identified due to its effects on migration of a fibroblast cell line), which has been implicated in the repulsion of neurites. As the authors note, in the CNS RTN4A has been previously associated with oligodendrocytes and myelin and not with astrocytes. The authors chose to examine RTN4A expression in astrocytes further. (c) RTNA4 expressing astrocyte nuclei also exhibited high levels of immune regulatory DEGs. (d) In vivo immunohistochemistry showed that marmoset astrocytes exhibited substantially and significantly higher levels of RTN4A protein compared with mouse astrocytes, in which expression was low. In vivo immunohistochemistry also confirmed RTN4A protein expression in human astrocytes near stroke tissue. The immunohistochemical images are of high quality and the results look specific and convincing. Overall, from a technical perspective, the work appears well conducted and properly controlled and the main findings appear to be reliable. Nevertheless, I have concerns regarding some of the interpretations and claims made on the basis of the findings.
Specific comments and concerns: 1. The authors claim to have identified a "primate-specific" astrocyte response to ischemia. However, they do not present convincing evidence to support such a strong claim as the main statement in the title of the paper. First, it is incorrect to say that RTN4 is not expressed by murine astrocytes. Various online databases show that uninjured astrocytes express significant levels of RTN4 (See the attached image from the online Barres lab database). The authors here themselves show that RTN4 is expressed also by reactive mouse astrocytes. Although this expression is relatively lower than in marmosets, what that means is not clear (nor is investigated). The authors are correct that this RTN4 expression in mouse has not been studied, but that does not mean it does not exist and might not be important in injury responses. The degree to which RTN4 does or does not contribute to immune regulatory functions of murine reactive astrocytes has not been studied, but that also does not mean that similar functions might exist in mouse as in marmoset. In the end, the authors only examine potential roles for astrocyte RTN4 in marmosets and then speculate that mouse astrocytes are fundamentally different because their level of RTN4 is somewhat lower. This is not sufficient to make a strong claim for a "primatespecific" astrocyte response that is presented in the title. The authors do not present any experiments that directly examine mouse astrocytes and macrophages and show that they behave differently from marmoset ones. At best the authors could discuss this as a possibility in their Discussion without claiming that this has been experimentally demonstrated. The main finding of the paper is that RTN4A (NOGOA) is expressed by astrocytes and limits peripheral macrophage infiltration after ischemic stroke in a non-human primate. The title should reflect this specific observation rather than try make a broad claim about species differences that have not been rigorously or convincingly shown. The title and text and discussion should focus on the specific findings that are made regarding RTN4, which are interesting, unexpected and important and merit publication in their own right. With an appropriate title, I think the paper would be appropriate for this journal.

A minor point:
The Discussion starts off with the statement that they provide the first dataset on single cell transcriptome changes in NHP astrocytes after brain injury. Quite frankly, claims of priority of demonstration like this are at best tedious. Other high profile journals like expressly PNAS forbid them. Such claims are best omitted.

Reviewer #1 (Remarks to the Author):
Studies in primates are often of crucial value on the way of a project or paradigm toward clinical translation. Here, Boghdadi et al. characterize astrocyte responses in visual cortex gray matter after a local ischemic stroke in marmosets by profiling their transcriptome 7 d. after injury. A number of plasticity and inflammation related genes and gene networks were found to be upregulated, among which the membrane protein Nogo-A (also in human strokes, but much less so in rodents!). Via the known Nogo-A receptor LILRB2/PirB Nogo-A restricts macrophage migration into the peri-infarct parenchyma as shown by knock-out experiments. All these observations have relevance for our understanding of the degenerative and repair processes at sites of CNS tissue destruction e.g. after stroke, as well as for the planned and currently ongoing clinical trials with anti-Nogo-A therapeutic agents. They thus address a basic neuroscience audience as well as clinical neuroscientists.
1. Where precisely in the visual cortex where the astrocyte nuclei for the transcriptomic analysis sampled: All layers of the gray matter? How close to the stroke core? How restricted to the periinfarct ischemic area/penumbra? Our apologies that this was not clear in the initial submission. Please refer to Figure 1a (line 136), which has been modified to clarify that the grey matter, encompassing all cortical layers of V1 operculum, including the stroke core and penumbra, was sampled. Further we have added more detail to the Methods (lines 571-583) and Results (lines 75-76). 3. Fig. 3 shows a comparison of mouse and marmoset astrocytes. The primate astrocytes appear much larger in size, in addition to being strongly Nogo-A-positive. Is this typical and real or a sampling artifact?
Yes, this is normal. Primate astrocytes are much larger in size when compared to mouse astrocytes. See data below from a recent publication of our lab. This observation was added to Results (lines 230-232). 5. The observation that also the Nogo-A fragment D20, which binds to receptors other than NgR1 or PirB/LILRB2, inhibits macrophage adhesion/migration is very interesting. The discussion of this should take into account that very little is known on the detailed composition of the Nogo-A receptor complexes. In analogy to e.g. neurotrophin or Wnt receptors, there is a good chance that different subunits form a multisubunit heteromeric complex. Thus, antibodies against one subunit could easily sterically hinder the binding of other subunits to a different ligand site (e.g. anti-LILRB2 could block binding of d20 to S1PR2 or syndecans). Integrins have only been shown to be indirectly involved. The scheme shown in Fig.5d is misleading. Figure 5d (line 392) has been modified to include both NgR1 and S1PR2 but are grayed out given we have demonstrated their absence on human macrophages. The schematic in Figure  5d (line 392) is the simplest model and it is possible that other NogoA receptors could be involved and be directly or indirectly inhibited by antibodies as a result of cross-subunit steric effects.
This point has been added to the Results (lines 378-384).
6. Macrophages have multiple functions in CNS lesions, 'good' and 'bad' ones. Effects of restricting their access to the CNS will depend on the time point after injury, the subpopulation of macrophages etc.
We completely agree with this comment. As highlighted in the Discussion (lines 509-510), macrophages are crucial for the clearance of detrimental cell and myelin debris after CNS injury. Astrocyte corralling of peripheral macrophages is an essential function following ischemia, evidenced by experimental data demonstrating worse functional outcomes when abolishing this function (lines 511-513). We hypothesize that astrocyte corralling at 7 DPI is a positive interaction, keeping the macrophages where they need to be to clean up and stabilize the injury site. In the acute stages post-ischemic stroke, peripheral macrophages are a mix of pro-and anti-inflammatory phenotypes, with anti-inflammatory phenotypes dominating (1-3). However, over time, there is a downregulation of anti-inflammatory genes whereas pro-inflammatory genes persist (1-3). Akin to others (4), we do not think limiting peripheral macrophage infiltration is a viable strategy following stroke. Ideally, we would want to find a way to limit the persistence of a pro-inflammatory phenotype more chronically.

Reviewer #2 (Remarks to the Author):
This manuscript from Boghdadi and colleagues investigates the heterogeneity of astrocyte responses in a marmoset model of ischemic stroke, as well as highlighting some mechanistic studies around the putative function of Nogo-A in mediating this response by blocking entrance of peripheral macrophages into the brain following insult. This work represents, to my knowledge, the first such analysis of cell-type specific transcritpomic changes in the marmoset model. Indeed the hypothesis that 'immediate post-stroke blockade of NogoA action may exacerbate brain inflammation' is intriguing, and adds a new layer of complexity to ongoing therapeutic trials using anti-Nogo therapies in spinal cord injury -these new data add a great deal to the literature surrounding Nogo as a purely inhibitory molecule. The authors should also be commended for taking some time to compare key findings across marmoset, rodent, and human -particularly important given their affirmation that the marmoset model provides a more appropriate model of the human ischemic response.
I do however have a few concerns with the data as presented in the manuscript, and would like clarifications or additional details to determine how they would alter the conclusions drawn from the study: 1. single nuclei RNASeq analysis -the assertion that astrocytes from stroke and control brains are distinctly different at the transcriptomic level is big surprise and may suggest the data was not anchored or integrated properly. Where there absolutely no physiological (ie. normal/control) astrocytes left following ischemic stroke? Even though V1 visual cortex was dissected out for downstream nuclei sequencing, it still seems unusual that a such a large region was collected that there are no baseline/normal astrocytes left. There needs to be more comprehensive descriptions of what analysis packages/pipelines were used to generate these final data -for instance, it seems likely that the different datasets (control/ischemia) have not been anchored properly (as stated above) -leading to an artifact of complete dataset separation. Canonical Correlation Analysis (or similar) should be used to anchor the experimental groups based on multiple genes that are shown not to change in individual cell types (e.g. for astrocytes perhaps Aldh1l1, Sox9, etc.). It would also be important to include additional QC data pertaining to gene, UMI, etc counts, and highlight how individual animal samples cluster across these small datasets -it is becoming apparent in other published under-powered datasets that the often individual samples are driving much of the reported heterogeneity and the authors would benefit from showing this is not the case with their own datasets. It is also unacceptable that raw data is not deposited at a public repository such as the NIH GEO.
The transcriptomic data has been re-integrated using an additional 1000 anchor points, which revealed greater overlap. This does not change any of the downstream data analysis, but only the UMAP & feature plots in Figure 1d  The data has been added to a public repository (NCBI GEO) and following statement added to the "Data and materials availability" section of the manuscript "All sequencing data that support the findings of this study have been deposited in the National Center for Biotechnology Information Gene Expression Omnibus (NCBI GEO) and are accessible through the GEO Series accession number GSE179141. All other relevant data are available from the corresponding author on request." (Lines 1081-1085).
To review GEO accession GSE179141: Go to https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE179141 Enter token mpsrqmgapxehfmn into the box 2. additional to this, the samples sizes are quite low, however given the difficulty of obtaining NHP samples this should be considered in that context. The authors should discuss this limitation. Indeed, the astrocyte capture in both control and ischemic animals is among the lowest of all cell types sequenced (~5%), while the capture rate for non-astrocyte cells, particularly of oligodendrocytes that express a lot of RTN4 (the gene that encodes Nogo-A), is much higher. It was a shame not to see these data also integrated into the manuscript -as they are much better powered for making some of the strong conclusions being drawn here.
While our sample size is low, and this is an important consideration with nonhuman primate tissue, our QC plots demonstrate we have collected enough nuclei from each individual marmoset, and we do not see any obvious differences among them.
We completely agree that looking at oligodendrocytes, the cell type previously associated with NogoA, would indeed be interesting. However, the focus of this particular paper is astrocytes and the unique finding that NogoA is upregulated on them following ischemic stroke. Rest assured that future research by the groups involved in this current study is underway for nonastrocyte cell types.
3. Fig 1 -RTLN4+ cell gene expression -GAP43 is also enriched in RTN4-astrocytes. Can the authors comment on it's inclusion here as an RTN4+ marker? GAP43, KLF6, CD44 were all simply chosen as markers for in tissue validation of the transcriptomic data. We are not claiming that GAP43 is a marker of RTN4+ astrocytes. GAP43 was differentially expressed within the top 100 genes for both RTN4-and RTN4+ astrocyte after injury compared to control. These genes were also chosen partially due to the HumanBase functional analysis performed earlier. GAP43 being expressed on both RTN4+ and RTNpopulations gave us a better spread for the cell counts between control and injury, which is why we chose to quantify GAP43 specifically. This was performed to demonstrate the reliability of our transcriptomic data to our in-tissue data. For example, the transcriptomic data tells us that 51% RTN4+ astrocytes express GAP43 in the injured cohort compared to 6% in the control cohort. Our in-tissue validation counts revealed very similar numbers: 44% in injured and 7% in control. experiment was to show that LILRB2+ human macrophages were repelled by NogoA. The fact that mouse and marmoset data was presented here was simply to show that NogoA was expressed by them. Additionally, as suggested by the reviewers of our manuscript, all references to primate specificity have been removed from the manuscript, so the treatment of human astrocytes with IL6 is no longer relevant. 5. TMEM119 and microglia (line 255) -TMEM119 levels have been reported to drop in microglia in the context of infection, injury, and disease. Can the authors comment on this as an alternate interpretation of the TMEM119-IBA1+ cells they labeled here. Given the authors have nucSeq data on a number of microglia (up to 26% of all collected nuclei in the stroke setting) they could mine these data to see if there are any Iba1+ cells in this population that do not express TMEM119.
Although some studies have reported a drop in TMEM119 levels in the context of infection, injury and disease, studies also support using TMEM119 as a marker of reactive microglia in human traumatic brain injury (1), in human multiple sclerosis grey matter lesions (2), and in injured mouse spinal cord (3).
Unfortunately, TMEM119 is not annotated in our marmoset genome alignment. Additionally, within our defined microglia (macrophage) cluster there also exists a large proportion of peripheral macrophages. Therefore, these 2 factors combined means that we were unable to look for Iba1+/ TMEM119-nuclei in our transcriptomic dataset and answer this question. Therefore, we cannot exclude that there are TMEM119-microglia at the infarct site.
A subset of this point has been added to Results (lines 262-264).