Review of miR-92a and miR-125a regulate key features of Eomes+ Tr1-like cells

General Comments
This is a well-written study on an area worthy of investigation, being the miRNome of human Tr1 cells to identify unique miRNAs relevant to Tr1 cell biology. The authors chose to focus on miRNAs that had the lowest expression in Tr1 cells, selecting miR-125a and miR-92a. Their claims are summarised in Fig 4F, that miR-92a suppresses EOMES downstream IFNγ and GZMK expression and that miR-125a suppresses IFNγ and IL-10RA. However, the data presented are not robust enough to support these claims. Although the luciferase reporter assay with a dual-transfected cell line showed some effect of miR-92a on EOMES 3’UTR levels, transfection of miR-92a into primary cells had almost no effect on EOMES expression. Transfection of miR-125a did not affect IL-10RA or IFNγ levels.

I have suggested some changes to the data analysis and would urge the authors to rethink the interpretation of their data. They have novel data that will reveal unique miRNAs in Tr1-like cells and their data suggest there could be a functional role. However, these roles may not be related to the genes they have focused on or may be differentially regulated in vivo. In the absence of additional data, I suggest a reframing of the paper with less emphasis on the function of the identified DE miRNAs without robust supporting data.

Specific Comments

1. How were the 424 differentially expressed genes determined (Table 1)? Were these genes that were differentially expressed when comparing Tr1-like cells to each of the other three subsets? How many were higher/lower in Tr1 cells? It is preferable to perform DE gene analysis of sequencing data using a differential gene expression package, such as DESeq2 or EdgeR, which enable comparisons across multiple groups (PMID: 26732976 shows a comparison to ANIVA). It would be useful to display the DE genes as a Venn diagram to show unique DE genes to Tr1-like cells in comparison to Th1EM and CTL cells. Can then list Tr1 unique DE genes into up and down regulated genes in a table.

2. Please convert all bar graphs to show all data points.

3. Data in Figure 4E show no significant differences between IL10RA and IFNγ at either protein or gene level and data for protein expression are from n=2. An objective assessment of these data would conclude that the miR-125a mimic has no effect on IL-10RA or IFNγ protein expression.

4. Most biological data is not normally distributed, and the data sets here are too small for a test of normality. Therefore, it is more robust to use non-parametric statistical analysis, for example a Kruskal-Wallis test rather than an ANOVA. The data metrics shown should also be non-parametric, for example, median and interquartile range, rather than mean and standard error or standard deviation.

5. In line with comment 5, p cannot be <0.05 with the large amount of overlap in the SEM bars in Figure 3E. This analysis should use a Mann-Whitney U test. It should be noted that such a small reduction in MFI is unlikely to be biologically significant.

Minor Comments
6. Table 1 is not easy to interpret; this large dataset could be supplemental or available through a data repository. More useful would be the list of DE genes that are the most significantly different in Tr1 cells to Th1EM and CTL cells in order of expression levels in Tr1-like cells.

1. Results page 4 “Compared to the other T-cell subsets investigated, miR-125a and miR-92a were expressed at the lowest levels in Tr1-like cells” needs rewording as Figure 2B shows miR-125a is not different between Tr1 and CTL cells and miR-92a is not different between Tr1 and Th1EM cells.

2. Similar to comment 7, cannot make a claim about inverse correlation with EOMES and miR-92a expression without (1) a correlation analysis and (2) when miR-92a levels in Tr1 cells are similar to Th1EM.

3. EOMES protein in capitals and non-italicised for humans and lowercase non-italicised for mouse.

4. Show representative FACS plots for all staining, including IL-10RA

5. Introduction states “Eomes expression in all other human CD4+ T-cell subsets is largely undetectable.” It is then stated that CCR5+CD4+ T cells sorted for Fig 3E are enriched for EOMES expression, please clarify, and show flow cytometry raw data for these expression levels. Why was effect of miR-92a on EOMES not investigated within Tr1-like cells isolated by defined surface markers in methods?

6. In discussion it is stated that “Tr1-like cells are likely to promote the persistence of viruses”, this ignores literature suggesting that Tr1 cells are crucial for generating anti-viral T cell memory (PMID: 26147684, PMID: 32424367).

7. What multiple testing correction was used for ANOVA analysis of miRNA expression data?

8. Statistical analysis performed need to be stated in every instance in all figure legends, a lot are missing,

9. Was the staining in Fig 4B done on unstimulated or activated cells?

General comments

The authors have done an excellent job in reworking the paper and addressing all previous comments. The manuscript is easy to read, the key points are clear and supported by data and the language has been adjusted appropriately when discussing the more preliminary data. In only have a few additional points to improve clarity. Congratulations on a lovely piece of work.

Major Point

1. When discussing/presenting data on DEGs need to specify whether you are discussing upregulated or downregulated genes or both. It would be useful to split the analysis and show two separate Venn Diagrams for up and down regulated DEGs in sFigure 1. It is also interesting to note that, even though the Tr1 to Th1CM comparison identified the highest number of total DEGs, only 28% of these DEGs were upregulated, and for the comparison with Th1EM only 33% of DEGs were upregulated. This contrasts with Tr1 comparison with CTL where 57% of DEGs were upregulated. This is worth highlighting. I think it is also interesting, and worth highlighting, that of the 8 DEGs in common between Tr1 comparisons with CTL and Th1CM, GZMK is the only one that is upregulated in Tr1 in both comparisons.

Minor Points

1. Define Tr1, Th1CM, Th1EM and CTL at first use.

2. sFigure1. In panel A, label the CD25+CD127low gate as Tregs so the next gate labelled as w/o Tregs is clear for readers not familiar with Treg gating. IL-7R should be IL-7RA (or CD127, which is more commonly used). Adjust font in panel C to use capitals where appropriate to match main text, if there are no DEGs common for all three comparisons then place a 0 in centre of Venn diagram.

3. If sTables can be uploaded as excel files then great, otherwise need to format for viewing in landscape mode in word/pdf. For both sTables, if not uploading as excel file then indicate in a footnote what text highlighting indicates, sTable 1 needs legend for this in excel file.

4. Would be useful to sort data in sTable 2 by up/down so reader can look separately at genes that are up and down regulated.

5. Be consistent with terminology. There is a mix of Tr1-like cells and Tr1-cells (in main text and figures/legends), same with IL-10R which is also a mixture with IL-10RA and IL-10Rα.

6. Page 2, line 47 “We confirmed elevated expression of EOMES, GZMK and GZMA in Tr1-cells”…compared to which population? All three genes only upregulated compared to Th1CM.

7. Page 2, line 52 ‘than Tr1-like cells’ (not then)

8. sFigure 2A, suggest an increase in font size for axes labels, hard to read.

9. Fig 2D use uncapped lines to indicate statistical comparisons, remove shadow from statistical comparison line in E. Write ‘scrambled’ in full on axes labels.

10. Figure 2 legend “E. Dual-luciferase assay in HEK-293T cells which transfected with”, should be ‘which were transfected with….’

11. sFigure 2E legend is missing phenotype definition of control “Histogram Overlay showing ex vivo IL-10R expression in gated Tr1-cells. The red line shows the negative Control (),”