Write a comment

This review reflects comments and contributions by Ryan Cubero, Femi Arogundade, Ana Dorrego-Rivas, Anushree Krishnamurthy, Anna Oliveras. Synthesized by Ryan Cubero

In this paper, Käseberg et al. took advantage of the clonal nature and the retention of the X-chromosome activation status in human induced pluripotent stem cells (hiPSCs) and combined cell reprogramming and single iPSC clone selection with in vitro neural differentiation protocols into NPCs and neurons or brain organoids and allele-specific analyses of bulk- and scRNA-seq data to show a tissue-, lineage- and developmental window-specific escapism of genes in the inactive X-chromosome during neural differentiation. This results in specific patterns of gene expression where some genes are expressed from only one of the two X-chromosomes (monoallelic expression) while others are expressed from both (biallelic expression). This novel description of differentiation-induced biallelic expression of reactivated genes is shown to create a neuroprotective bias in females, by selectively opening up the silenced allele pool to a transcriptionally active state as well as employing protein regulation machinery. This confers an advantage to females in the context of X-linked neurodevelopmental disorders, and offers a certain level of resilience against the manifestation of disease phenotypes during this developmental window. 

Overall, we believe that the paper’s strengths outlined below make their findings highly valuable to the scientific community:

• Employing clonal human female induced pluripotent stem cells (iPSCs) represents a key strength of this study, as it enables the tracking of X-chromosome inactivation states in individual cells throughout the course of neural differentiation. The methodology yields more accurate and dependable data compared to heterogeneous cell populations.

• The use of brain organoids as a model system is a valuable approach to study neural differentiation and the impact of X-chromosome reactivation on neural tissue resilience. Brain organoids can mimic certain aspects of human brain development, allowing for investigations into complex cellular interactions.

• The comparison of brain organoids derived from male and female iPSCs carrying mutations in the MID1 gene is an important aspect of the study. Such comparative experimental setup allows for investigating the influence of X-chromosome reactivation on the phenotypic presentation of X-linked neurodevelopmental disorders.

• The research substantiates X-chromosome reactivation through the application of allele-specific RT-PCR and Western blot analysis. This empirical validation enhances the support for the presence of reactivation within neural cells.

• The researchers employed a comprehensive set of experimental approaches, including allele-specific expression analyses, single-cell RNA sequencing (scRNA-seq), bulk RNA-sequencing, allele-specific RT-PCR, and Western blot analysis. The combination of these methods provides a robust foundation for the conclusions drawn in the study.

• Null (background) distributions used for statistical testing, where applicable, were estimated from the data by random sampling. We commend the authors for such approaches to perform significance testing.

Moreover, we do believe that the following aspects need to be addressed to strengthen the observations and interpretation made in the study, as well as improve the readability of the manuscript:

Comments on the bulk RNAseq experiment and analyses:

• When comparing M-line and J-line (Fig 1), the authors find a very limited amount of overlap (i.e., genes classified in the same class in both lines) and quite a lot of genes that are specific for each line. This suggests that, rather than a general and controlled mechanism, the expression landscape is cell line dependent. Authors should comment how this limited overlap and variability between cell lines fit with their hypothesis of a "fine-tuned mechanism of gene expression during female neural development".

• Is there a particular reason why the authors chose 0.025 as a cut-off to call the gene as reactivated? What about the 0.1 cut-off for full escapees?

• If the reactivated genes are spatially concentrated in chrX, it could also be assumed that they are as far from the late-silenced genes as possible. Furthermore, looking at the ideograms (Fig 1C and 1E), it does appear that reactivated genes are close to full escapees. Could these observations be strengthened? These cross-categorical distance analysis could potentially give hints regarding how these genes escape inactivation.

• Authors should introduce better the advantages of using fetal brain tissue for this analysis, how to compare it to iPSCs lines and which set of genes was analyzed/plotted.

Comments on scRNAseq experiment and analyses:

• For the organoid derivation (3D), the protocol starts from iPSCs, as well as for the (2D) differentiation of NPCs and further neurons. The authors should indicate at which day of derivation the organoids were used for sequencing. Do these time points overlap with the NPCs and neurons that the authors bulk sequenced? It would be interesting to compare the bulk RNA seq from NPCs and neurons with the clusters of neuronal progenitors and mature neurons found in the organoids. Moreover, if one performs a pseudo-bulk analysis on these scRNA clusters, will we recapitulate the bulk RNAseq results?

• Based on the scRNAseq, the brain organoids also contain mesenchymal and choroid plexus populations. Does SOX2 mark these cell types as well? Can the scRNAseq support this finding and potentially, also whether other cell types are overrepresented in the different conditions?

• What is the degree of agreement between these 26 X-linked genes with a traceable SNP in the scRNAseq with those found in the RNAseq? Are the expression levels from progenitor cluster to neuron in the scRNAseq in the same directionality as the bulk RNAseq?

• The authors wrote “It further draws the surprising picture that switching between mono- and biallelic expression of X-linked genes is a frequently used, fine-tuned mechanism of gene expression regulation during female neural development. ” In my opinion, the authors should expand on this conclusion. They have the full Fig 3 investigating the differential switching between mono- and biallelic expression along lineage trajectories in brain organoids. I think that they have enough strong data to support a tightly regulated mechanism of X-linked gene expression during female neural development.

Minor comments on scRNAseq experiment and analyses:

• How large is Cluster 11? The fact that this cluster lacks cells from J-OS/hom condition suggests that this might be interesting to dig deeper into.

• Why is the biallelic expression summed up instead of taking the average?

• The authors wrote “If a gene contained more than one type of biallelically expressed variants, the gene was assigned to a single category”. How many of such instances were there?

Comments on reporting:

• Incorporating a dedicated section within the preprint to delve into the study's limitations is imperative. A comprehensive discourse on these limitations is vital to offer a well-rounded interpretation of the presented results.

• Supplementary tables are missing in the preprint.

• We commend the authors for seeming to be open in terms of data sharing. It would be good as well if the authors would eventually upload their codes in a public repository.

• Optional for the end of the 2nd paragraph in the Introduction: I would suggest here rather a high-level than a detailed description of the results.

• Panels in Fig 1 (all) and panels F and G in Fig S5 are very small and difficult to read. I suggest the authors make those panels bigger, splitting into several figures if needed.

Moreover, overall the statistical significance in the graphs (asterisks, ‘n.s.’) is almost illegible. See examples in Fig S4.

• In Figs 1B,D and F: Are these types of data plots, scatter plots?

• In Figs 1B,D and F, Fig S2H, Fig S3B, and Figs S5A,B,C and D, it will be good to have the genes arranged alphabetically to facilitate cross-checking with other figures.

• In Figs 1C,E and G, I think the loci in magenta represent the centromere. I would suggest that the authors indicate it in the figure legend.

• Optional for Fig1: For a better understanding of Fig 1, I would include schematic Fig S1A explaining the logic of clone selection including the color code to read the following data

• Optional for Fig 1: For the uninitiated, it would be helpful to provide a background on the M-ctrl and J-ctrl cell lines.

• Optional for Fig 2F: The key to the grey-coloured clusters could be specified.

• In Fig 3, I think the numbers in each dot is cluster ID, which are sequentially attributed following the lineage trajectories. It will be good to have this indicated in the caption.

• In Figs 4C and D, it would be better if the range on y-axis is the same on both plots to easily compare M-OS/male and J-OS/hom.

• It would be helpful if Fig S2H contains information about whether the genes depicted are reactivated, full escapee or late-silenced genes (as in Figs 1B,D and F) and whether it was observed in either NPC, neurons or both. It will also be helpful if the groups of genes in Figs 1B,D and F will be arranged alphabetically to facilitate comparison between these datasets.

• In the Microscopy and image analysis section, the authors need to include more details in this section. Are all the acquisitions 2D, or are they stacks? Details about the chosen objective and pixel size need to be specified.

• Immunohistochemistry: Where applicable, information about antibodies should include their RRID (Resource Identification Portal) code.

• Whenever the authors use a function (during analysis) which requires parameters, it would be good to report the parameters used, or write down default parameters if no special parameter choice was done. It will also be good to have non-default parameters noted down, if any. Else, it can be written "with default parameters".

• The Statistics and reproducibility section in the methods needs details about if and which tests were performed to assess data normality.

Suggestions for future studies

• The authors found a very limited overlap of genes when comparing M-line and J-line (Figure 1), a variability that could suggest that, rather than a general and controlled mechanism, the expression landscape is cell line dependent. An analysis of more than 2 iPSC lines can help to understand such variability.

• I wonder if the authors have any data comparing the epigenetic landscape of reactivated genes versus late-silenced genes? The authors could use gene regulatory network inference algorithms (e.g., SCENIC for scRNAseq or ISMARA for bulk RNAseq) for this.

We wish the authors all the best on continuing this line of research!

Competing interests

The author declares that they have no competing interests.