Molecular Consequences of Peripheral Influenza A Infection on Cell Populations in the Murine Hypothalamus

Reviewer #1 (Public Review):

This study uses single-cell genomics and gene pathway analysis to characterize the transcriptional effects of influenza H1N1 infection on cell types of the lateral hypothalamus and dorsomedial hypothalamus. The authors use droplet-based single-nuclei RNA-seq to profile single-cell gene expression at 3, 7, and 23 days post intranasal infection with H1N1 influenza virus. Through state-of-the-art and rigorous computational methods, the authors find that many hypothalamic cell types, including glia and neurons, are transcriptionally altered by respiratory infection with a non-neurotropic influenza virus, and that these alterations can persist for weeks and potentially affect cell type interactions that disrupt function. Their thorough discussion of the findings raises interesting questions and hypotheses about the functional implications of the molecular changes they observed, including the physiological changes that can persist long after acute viral infection. Given the role of the hypothalamus in homeostasis, this work sheds light on potential mechanisms by which the H1N1 virus can disrupt cell function and organismal homeostasis beyond the cells that it directly infects.

Despite its strengths, there are several points in the manuscript lacking sufficient evidence or clarity, which need to be addressed through revision. For instance, the conclusion that neurons but not non-neurons show persistent changes in gene expression may be alternatively explained by differences in the number of neuron and non-neuronal cells and transcripts. Also, the authors highlight the connection between influenza infection and loss of appetite and sleepiness but do not explore whether the influenza infection affected the cell types in their dataset previously associated with appetite and sleepiness, or whether differences in weight loss among the influenza-infected subjects correspond to any differences in gene expression.

Reviewer #2 (Public Review):

The new work from Lemcke et al suggests that the infection with Influenza A virus causes such flu symptoms as sleepiness and loss of appetite through the direct action on the responsible brain region, the hypothalamus. To test this idea, the authors performed single-nucleus RNA sequencing of the mouse hypothalamus in controlled experimental conditions (0, 3, 7, and 23 days after intranasal infection) and analyzed changes in the gene expression in the specific cell populations. The key results are promising.

However, the analysis (cell type annotation, integration, group comparison) is not optimal and incomplete and, therefore should be significantly improved.

More specifically:

1. The current annotation of cell types (especially neuronal but also applicable to the group of heterogeneous "Unassigned cells") did not make a good link to existing cell heterogeneity in the hypothalamus identified with scRNA seq in about 20 recently published works. All information about different peptidergic groups can not be extracted from the current version (except for a few). There are also some mistakes or wrong interpretations (eg, authors assigned hypothalamic dopamine cells to the glutamatergic group, which is not true). This state is feasible to improve (and should be improved) with already existing data.

2. I am confused with the results shown in the label transfer (suppl fig 3 and 4; note, they do not have the references in the text) applied to some published datasets (authors used the Seurat functions 'FindTransferAnchors' and 'TransferData'). The final results don't make sense: while the dataset for the arcuate nucleus (Campbel et al) well covered the GABAergic neurons it is not the case for the whole hypothalamus datasets (Chen et al; Zeisel et al). Similarly, for glutamatergic neurons. Additionally, I could not see that the label transfer works well for PMCH cells which should be present in the dataset for the lateral hypothalamus (Mickelsen et al,2019).

3. There are newly developed approaches to check the shifts in the cell compositions and specific differential gene expression in the cell groups (e.g. Cacoa from Kharchenko lab, scCoda from Büttner et al; etc). Therefore, I did not fully understand why here the authors used the pseudo-bulk approaches for the data analysis (having such a valuable dataset with multiple hashed samples for each timepoint). Therefore it would be great to use at least one of those approaches, which were developed specifically for the scRNAseq data analysis. Or, if there are some reasons - the authors should argue why their approach is optimal

4. When the authors describe the DGE changes upon experimental conditions (Figures 5 and 6), my first comment is again relevant: it is difficult to use the current annotation and cell type description as the reference for testing virus effects and shifts in the DGE in distinct neuronal subtypes.

I have to note that the experimental design is well done and logical. Therefore I believe that to strengthen the conclusions, the already obtained datasets can be used for improved analysis.