Écrire un commentaire

Summary

A central question in cancer biology is the aneuploidy paradox: although aneuploidy is typically harmful to normal cells, it is associated with increased proliferation in cancer. In this manuscript, the authors investigate this paradox by utilizing single-cell RNA-seq data from the Curated Cancer Cell Atlas. By generating matched gene expression and inferred aneuploidy profiles, the authors examine the impact of aneuploidy in tumors at single-cell resolution. Interestingly, they observe opposite patterns in their single-cell and bulk-tumor analyses with respect to proliferation phenotypes. When comparing cells with a high aneuploidy score (AS) to those with a low AS in single-cell data, aneuploidy is associated with reduced proliferation and metabolism; however, when comparing high AS tumors to low AS tumors in bulk data, aneuploidy is associated with increased proliferation and metabolic activity. The authors effectively demonstrate that tumors with low AS tend to be more heterogenous than tumors with high AS, supporting a model in which newly emerging aneuploidies create cellular stress and decreased proliferation while cells carrying advantageous alterations survive and expand. Recurrent chromosome-arm alterations were associated with consistent gene expression changes: loss of Chr9p was associated with EMT pathway upregulation, loss of chr18q was associated with reduced TGF-b signaling, gain of Chr8q was associated with MYC upregulation, and gain of Chr1q was associated with increased Notch signaling. This confirms previous work from the literature that utilized bulk sequencing analysis. Highly aneuploid tumors were depleted for T cells and B cells and enriched for M2-like macrophages, suggesting a distinct immune microenvironment compared to low aneuploid tumors.

This work is important as it addresses the aneuploidy paradox, which is a key question in the field, at single-cell resolution. The single cell approach is valuable because it allows for insights about aneuploidy that cannot be gleaned with bulk tumor-level analyses. The contrasting patterns for proliferation phenotypes observed in the single cell and bulk analyses are important to note when interpreting other studies of aneuploidy in tumors. Some of the other aneuploidy-associated phenotypes appear robust regardless of sequencing approach, as the authors showed similar bulk-level patterns in TCGA data and tested methods of various stringency. While re-analysis of these published single cell RNAseq datasets is valuable, the results would be strengthened by independently-derived datasets, as well as better controlling for tumor type and cell-type composition. Additionally, some conclusions about specific pathways (such as p53) could be considered overstated due to weak correlations - these concerns could be mitigated by additional text on the limitations of the study.

Major Points:

1. Despite the non-uniform representation of cancer types in the pan-cancer cohort, several of the initial analyses do not account for tumor type. Thus, observations from analyses such as GSEA could be disproportionately driven by the effects of aneuploidy in particular tumor types. Although Supplementary Figure S3 includes intra-tumor analyses, the individual tumors are not labelled, making it difficult to determine how tumor type may be impacting the results. Additionally, individual tumor samples have vastly different numbers of cells; samples with more cells could disproportionately influence the results. The authors could consider and clarify whether and how tumor type and differences in sample-level cell counts could impact the reported results.

2. The aneuploidy score provides a general measure of aneuploidy but does not distinguish between the effects of specific chromosomal gains and losses. The authors could consider clarifying this limitation and evaluating whether the reported associations are consistent across different types of aneuploidy. This would help evaluate whether analyses that only consider AS reflect the general effects of aneuploidy, or whether they are capturing specific aneuploidy patterns.

3. Inferring copy number from scRNA-seq, then analyzing the effects of copy number within the same transcriptomic data, could be considered circular. Thus, incorporation of additional existing datasets that derived both genomic DNA profiles and transcriptomes from single cells would be beneficial to strengthen the authors' claims.

4. The supplementary tables referenced in the manuscript have not been made available. The authors could consider ensuring that all supplementary tables are accessible so that others are better able to utilize the data.

5. In Figure 4A, the authors evaluate concordance between bulk and single-cell analyses at the level of REACTOME pathways, but they do not perform a systematic gene-level comparison. Positive or negative correlations between pathway-level results provide limited information about the extent to which individual genes or which specific pathways exhibit concordant changes across the two types of analysis. The manuscript does include select gene-level examples, but the authors could consider including a systematic comparison of gene-level effect estimates and the results from each individual pathway analysis to provide a more complete assessment of the concordance between the analyses.

Minor Comments

1. The abstract does not mention the findings related to immune phenotypes. The authors could consider including these findings to provide a more thorough summary of the study.

2. The meaning of the outer color coding in Figure 1A is unclear. Additionally, the statement “number of source publications for each type is indicated in parentheses” is potentially confusing, as the numbers refer to the number of samples rather than the number of publications. The authors could consider altering the wording to “number of samples for each type” to clarify this.

3. Because AS is determined at the single-cell level before being averaged to obtain tumor-level values, the authors could consider switching Figures 1B and 1C to improve the logical flow of the figures.

4. It is unclear whether the differential gene expression analysis outlined in Figure 1D treats AS as a categorical variable (comparing low AS cells to high AS cells), or if it is a continuous variable with the numerical AS values. The authors could consider clarifying which approach was used and justify this decision.

5. The pathway labels in figures 2C and 2E could be made easier to read. Since the pathways are already stated to be Hallmark gene sets, using “Hallmark” at the beginning of each label is not necessary, and removing it would shorten the labels and thus improve readability. The authors could also reduce the number of labelled points and/or use colors to distinguish different functional categories that are mentioned in the text, such as proliferation and immune response.

6. The correlation involving p53 appears weak and does not provide particularly convincing support for the proposed association. Although the authors do mention that this correlation is weak, its significance is overstated. The authors could consider softening the claims involving p53 to better reflect the strength of the evidence.

7. The expression units and scale for CDKN2A expression in 4E is not clear, and the reported values seem low. The authors could consider clarifying the expression scale and ensure that these expression levels are biologically meaningful.

8. Because CDKN2A is located on Chr9p, its reduced expression in cells that have lost Chr9p does not seem particularly informative. The authors could consider clarifying the significance of this finding and what it means about the loss of Chr9p so that readers can better understand its importance.

9. The interpretation of the CDKN2A expression differences in Figure 4E could be influenced by cell cycle-associated effects. Since CDKN2A is involved in cell-cycle regulation, the authors could consider examining whether cell cycle-state differences could impact this comparison and how it may affect the interpretation of the results; this could help readers better understand the significance of these expression differences.

10. Some of the figure text (ex: cell types in Figure 5A and cancer types in Figure 5B) is too small. The authors could increase font sizes where necessary to improve readability.

I reviewed this preprint as part of the Transparent and Accountable Peer Review training program. This report was written with the support of my research mentor.

Competing interests

A member of my current research group collaborates with some of the authors on this preprint on an unrelated project. However, I am not involved in any such collaboration.

Use of Artificial Intelligence (AI)

The author declares that they did not use generative AI to come up with new ideas for their review.