Single-cell transcriptomics of X-ray irradiated Drosophila wing discs reveals heterogeneity related to cell-cycle status and cell location

Reviewer #1 (Public review):

Summary:

The authors analyze transcription in single cells before and after 4000 rads of ionizing radiation. They use Seuratv5 for their analyses, which allows them to show that most of the genes cluster along the proximal-distal axis. Due to the high heterogeneity in the transcripts, they use the Herfindahl-Hirschman index (HHI) from Economics, which measures market concentration. Using the HHI, they find that genes involved in several processes (like cell death, response to ROS, DNA damage response (DDR)) are relatively similar across clusters. However, ligands activating the JAK/STAT, Pvr, and JNK pathways and transcription factors Ets21C and dysf are upregulated regionally. The JAK/STAT ligands Upd1,2,3 require p53 for their upregulation after irradiation, but the normal expression of Upd1 in unirradiated discs is p53-independent. This analysis also identified a cluster of cells that expressed tribbles, encoding a factor that downregulates mitosis-promoting String and Twine, that appears to be G2/M arrested and expressed numerous genes involved in apoptosis, DDR, the aforementioned ligands, and TFs. As such, the tribbles-high cluster contains much of the heterogeneity.

Strengths:

(1) The authors have used robust methods for rearing Drosophila larvae, irradiating wing discs, and analyzing the data with Seurat v5 and HHI.

(2) These data will be informative for the field.

(3) Most of the data is well-presented.

(4) The literature is appropriately cited.

Weaknesses:

(1) The data in Figure 1 are single-image representations. I assume that counting the number of nuclei that are positive for these markers is difficult, but it would be good to get a sense of how representative these images are and how many discs were analyzed for each condition in B-M.

(2) Some of the figures are unclear.

Reviewer #2 (Public review):

This manuscript investigates the question of cellular heterogeneity using the response of Drosophila wing imaginal discs to ionizing radiation as a model system. A key advance here is the focus on quantitatively expressing various measures of heterogeneity, leveraging single-cell RNAseq approaches. To achieve this goal, the manuscript creatively uses a metric from the social sciences called the HHI to quantify the spatial heterogeneity of expression of individual genes across the identified cell clusters. Inter- and intra-regional levels of heterogeneity are revealed. Some highlights include the identification of spatial heterogeneity in the expression of ligands and transcription factors after IR. Expression of some of these genes shows dependence on p53. An intriguing finding, made possible by using an alternative clustering method focusing on cell cycle progression, was the identification of a high-trbl subset of cells characterized by concordant expression of multiple apoptosis, DNA damage repair, ROS-related genes, certain ligands, and transcription factors, collectively representing HIX genes. This high-trbl set of cells may correspond to an IR-induced G2/M arrested cell state.

Overall, the data presented in the manuscript are of high quality but are largely descriptive. This study is therefore perceived as a resource that can serve as an inspiration for the field to carry out follow-up experiments.

Reviewer #3 (Public review):

Summary:

Cruz and colleagues report a single-cell RNA sequencing analysis of irradiated Drosophila larval wing discs. This is a pioneering study because prior analyses used bulk RNAseq analysis, so differences at single-cell resolution were not discernible. To quantify heterogeneity in gene expression, the authors make clever use of a metric used to study market concentration, the Herfindahl-Hirschman Index. They make several important observations, including region-specific gene expression coupled with heterogeneity within each region and the identification of a cell population (high Trbl) that seems disproportionately responsible for radiation-induced gene expression.

Strengths:

Overall, the manuscript makes a compelling case for heterogeneity in gene expression changes that occur in response to uniform induction of damage by X-rays in a single-layer epithelium. This is an important finding that would be of interest to researchers in the field of DNA damage responses, regeneration, and development.

Weaknesses:

This work would be more useful to the field if the authors could provide a more comprehensive discussion of both the impact and the limitations of their findings, as explained below.

Propidium iodide staining was used as a quality control step to exclude cells with a compromised cell membrane. But this would exclude dead/dying cells that result from irradiation. What fraction of the total do these cells represent? Based on the literature, including works cited by the authors, up to 85% of cells die at 4000R, but this likely happens over a longer period than 4 hours after irradiation. Even if only half of the 85% are PI-positive by 4 hr, this still removes about 40% of the cell population from analysis. The remaining cells that manage to stay alive (excluding PI) at 4 hours and included in the analysis may or may not be representative of the whole disc. More relevant time points that anticipate apoptosis at 4 hr may be 2 hr after irradiation, at which time pro-apoptotic gene expression peaks (Wichmann 2006). Can the authors rule out the possibility that there is heterogeneity in apoptosis gene expression, but cells with higher expression are dead by 4 hours, and what is left behind (and analyzed in this study) may be the ones with more uniform, lower expression? I am not asking the authors to redo the study with a shorter time point, but to incorporate the known schedule of events into their data interpretation.

If cluster 3 is G1/S, cluster 5 is late S/G2, and cluster 4 is G2/M, what are clusters 0, 1, and 2 that collectively account for more than half of the cells in the wing disc? Are the proportions of clusters 3, 4, and 5 in agreement with prior studies that used FACS to quantify wing disc cells according to cell cycle stage?

The EdU data in Figure 1 is very interesting, especially the persistence in the hinge. The authors speculate that this may be due to cells staying in S phase or performing a higher level of repair-related DNA synthesis. If so, wouldn't you expect 'High PCNA' cells to overlap with the hinge clusters in Figures 6G-G'? Again, no new experiments are needed. Just a more thorough discussion of the data.

Trbl/G2/M cluster shows Ets21C induction, while the pattern of Ets21C induction as detected by HCR in Figures 5H-I appears in localized clusters. I thought G2/M cells are not spatially confined. Are Ets21C+ cells in Figure 5 in G2/M? Can the overlap be confirmed, for example, by co-staining for Trbl or a G2/M marker with Ets21C?

Induction of dysf in some but not all discs is interesting. What were the proportions? Any possibility of a sex-linked induction that can be addressed by separating male and female larvae?

Author response:

We thank the reviewers for their comments and for their constructive suggestions. We intend to submit a revised manuscript where we address the comments made in the Public Reviews as well as in the Recommendations for the Authors.

One of our most interesting findings, as noted by the reviewers, was the discovery of a small subpopulation of cells likely arrested in G2 that accounts for a disproportionate amount of radiation-induced gene expression. In addition, to the responses indicated below, we are planning to include additional “wet lab” experiments in the revised manuscript that address the properties of this seemingly important subpopulation of cells.

Reviewer 1:

Strengths:

(1) The authors have used robust methods for rearing Drosophila larvae, irradiating wing discs, and analyzing the data with Seurat v5 and HHI.

(2) These data will be informative for the field.

(3) Most of the data is well-presented.

(4) The literature is appropriately cited.

Thank you for these comments

Weaknesses:

(1) The data in Figure 1 are single-image representations. I assume that counting the number of nuclei that are positive for these markers is difficult, but it would be good to get a sense of how representative these images are and how many discs were analyzed for each condition in B-M.

(2) Some of the figures are unclear.

In the revised manuscript, we will provide a more detailed quantitative analysis. For each condition, we analyzed 4 - 9 discs.

We assume that the reviewer in referring to panels in Figure 1. We will review these images and if necessary, repeat the experiments or choose alternative images that appear clearer.

Reviewer 2:

Overall, the data presented in the manuscript are of high quality but are largely descriptive. This study is therefore perceived as a resource that can serve as an inspiration for the field to carry out follow-up experiments.

We intend to include more “wet lab” experiments in our revised manuscript to address the identity and properties of the high-trbl cells that we have identified using the clustering approach based on cell-cycle gene expression.

Reviewer 3:

Strengths:

Overall, the manuscript makes a compelling case for heterogeneity in gene expression changes that occur in response to uniform induction of damage by X-rays in a single-layer epithelium. This is an important finding that would be of interest to researchers in the field of DNA damage responses, regeneration, and development.

Thank you.

Weaknesses:

This work would be more useful to the field if the authors could provide a more comprehensive discussion of both the impact and the limitations of their findings, as explained below.

Propidium iodide staining was used as a quality control step to exclude cells with a compromised cell membrane. But this would exclude dead/dying cells that result from irradiation. What fraction of the total do these cells represent? Based on the literature, including works cited by the authors, up to 85% of cells die at 4000R, but this likely happens over a longer period than 4 hours after irradiation. Even if only half of the 85% are PI-positive by 4 hr, this still removes about 40% of the cell population from analysis. The remaining cells that manage to stay alive (excluding PI) at 4 hours and included in the analysis may or may not be representative of the whole disc. More relevant time points that anticipate apoptosis at 4 hr may be 2 hr after irradiation, at which time pro-apoptotic gene expression peaks (Wichmann 2006). Can the authors rule out the possibility that there is heterogeneity in apoptosis gene expression, but cells with higher expression are dead by 4 hours, and what is left behind (and analyzed in this study) may be the ones with more uniform, lower expression? I am not asking the authors to redo the study with a shorter time point, but to incorporate the known schedule of events into their data interpretation.

We thank the reviewer for these important comments. The generation of single-cell RNAseq data from irradiated cells is tricky. Many cells have already died. Even those that do not incorporate propidium iodide are likely in early stages of apoptosis or are physiologically unhealthy and likely made it through our FACS filters. Indeed, in irradiated samples up to 57% of sequenced cells were not included in our analysis since their RNA content seemed to be of low quality. It is therefore likely that our data are biased towards cells that are less damaged. As advised by the reviewer, we will include a clearer discussion of these issues as well as the time course of events and how our analysis captures RNA levels only at a single time point.

If cluster 3 is G1/S, cluster 5 is late S/G2, and cluster 4 is G2/M, what are clusters 0, 1, and 2 that collectively account for more than half of the cells in the wing disc? Are the proportions of clusters 3, 4, and 5 in agreement with prior studies that used FACS to quantify wing disc cells according to cell cycle stage?

Clusters 0, 1, and 2 likely contain cells in other stages of the cell cycle, including early G1. Other studies indicate that more than 70% of cells are expected to have a 4C DNA content 4 h after irradiation at 4000 Rad. The high-trbl cluster only accounts for 18% of cells. Thus clusters 0, 1 and 2 could potentially contain other populations that also have a 4C DNA content. Importantly, similar proportions of cells in these clusters are also observed in unirradiated discs. We are mining the gene expression patterns in these clusters with the goal of estimating their location in the cell cycle and will include those data in the revised manuscript.

The EdU data in Figure 1 is very interesting, especially the persistence in the hinge. The authors speculate that this may be due to cells staying in S phase or performing a higher level of repair-related DNA synthesis. If so, wouldn't you expect 'High PCNA' cells to overlap with the hinge clusters in Figures 6G-G'? Again, no new experiments are needed. Just a more thorough discussion of the data.

We have found that the locations of elevated PCNA expression do not always correlate with the location of EdU incorporation either by examining scRNA-seq data or by using HCR to detect PCNA. PCNA expression is far more widespread. We intend to present additional data that address this point and also a more thorough discussion in the revised manuscript.

Trbl/G2/M cluster shows Ets21C induction, while the pattern of Ets21C induction as detected by HCR in Figures 5H-I appears in localized clusters. I thought G2/M cells are not spatially confined. Are Ets21C+ cells in Figure 5 in G2/M? Can the overlap be confirmed, for example, by co-staining for Trbl or a G2/M marker with Ets21C?

The data show that the high_-trbl_ cells are higher in Ets21C transcripts relative to other cell-cycle-based clusters after irradiation. This does not imply that high-trbl-cells in all regions of the disc upregulate Ets21C equally. Ets21C expression is likely heterogeneous in both ways – by location in the disc and by cell-cycle state. We will attempt to look for co-localization as suggested by the reviewer.

Induction of dysf in some but not all discs is interesting. What were the proportions? Any possibility of a sex-linked induction that can be addressed by separating male and female larvae?

We can separate the cells in our dataset into male and female cells by expression of lncRNA:roX1/2. When we do this, we see X-ray induced dysf expressed similarly in both male and female cells. We think that it is therefore unlikely that this difference in expression can be attributed to cell sex. We are investigating other possibilities such as the maturity of discs.