The anatomy of transcriptionally active chromatin loops in Drosophila primary spermatocytes using super-resolution microscopy

While the biochemistry of gene transcription has been well studied, our understanding of how this process is organised in 3D within the intact nucleus is less well understood. Here we investigate the structure of actively transcribed chromatin and the architecture of its interaction with active RNA polymerase. For this analysis, we have used super-resolution microscopy to image the Drosophila melanogaster Y loops which represent huge, several megabases long, single transcription units. The Y loops provide a particularly amenable model system for transcriptionally active chromatin. We find that, although these transcribed loops are decondensed they are not organised as extended 10nm fibres, but rather they largely consist of chains of nucleosome clusters. The average width of each cluster is around 50nm. We find that foci of active RNA polymerase are generally located off the main fibre axis on the periphery of the nucleosome clusters. Foci of RNA polymerase and nascent transcripts are distributed around the Y loops rather than being clustered in individual transcription factories. However, as the RNA polymerase foci are considerably less prevalent than the nucleosome clusters, the organisation of this active chromatin into chains of nucleosome clusters is unlikely to be determined by the activity of the polymerases transcribing the Y loops. These results provide a foundation for understanding the topological relationship between chromatin and the process of gene transcription.

In this paper we exploit the model system of the Y loops in Drosophila spermatocytes to provide a super-resolution view of the topology of active transcription. We are pleased that Reviewer 1 assesses the work as "highly significant". They also mention the caveat that findings may be "specific to the model system". Reviewer 2 provides a counter to this caveat by pointing out that our findings are in line with some previous reports in mammalian cells. The interpretation in previous studies has however been hindered by the density of chromatin in the nucleus making it difficult to clearly identify individual fibres and Reviewer 2 points out that our work provides a significant advance as in our system we are able to "visualize the organization of chromatin within an individual, isolated loop".

Reviewer #1 • Can you stimulate and inhibit transcription to see how changes in RNA Polymerase and nucleosome organisation occur?
We are not aware of an approach to further stimulate transcription in this system, however we have some preliminary analyses of the effects of transcription inhibition that we plan to firm up within weeks. We agree with the reviewer that it may potentially be interesting and informative to include such studies but we do not consider this as central to the significance of the paper. So our plan is to attempt to quickly gather data on this for potential inclusion but we would plan to only include this material if we can rapidly achieve a clear understanding of the effects of transcription inhibition on nucleosome cluster organisation. Although data was collected with variable number of frames (20K-50K), in the original paper we only used data from 50K samples for quantitative analysis. In response to reviewer #2, we have now performed the analysis on a larger data set with variable frame number. Analysis of the two data sets controls for any impact of frame number variation and makes very little difference to the results of the analysis e.g the median cluster FWHM remains at 52 nm. We have now specified the details of our estimation of the number of nucleosomes per cluster in the Materials and Methods. Techniques to quantify clusters based on localisation number have been used in the past on SMLM data, but the normalisations are difficult to validate as there is uncertainty around numbers of antibodies binding, differences in frame numbers affecting blink number, as well as unknown binding efficiencies of antibodies. Our cluster quantification instead is based on the size of the clusters (cluster width), as this does not vary significantly across frame number once signal saturation has been achieved and avoids the difficulties associated with normalisation of localisation number. As we have mentioned in the Discussion, our "simple volume calculation" does have some caveats arising from the STORM precision, the labelling method and the EM-derived estimate of packing density; nevertheless, as we say in the Discussion we consider it provides a useful "starting point for interpretation of the observed labelling in terms of underlying chromatin structure."

Reviewer
• Methods, STORM imaging: the authors should provide the imaging buffer composition as the papers they refer to make use of an array of buffers. If possible, the authors should include laser powers in standard units (kW/cm^2) to increase clarity and reproducibility.
We have provided the imaging buffer composition and the laser power in standard units.

• Methods, Cluster analysis: the authors should specify the values for the parameters used in
"First a description of the average cluster was established using spatial statistics, then a clustering algorithm using the average parameters provided by the initial spatial statistics further refined the description" As detailed in response to the next comment, we have clarified the text in the Materials and Methods to specify the parameters and workflow of the analysis.
• Given the availability of broadly used methods for clustering such as DBScan (Ester, et al. A density-based algorithm for discovering clusters in large spatial databases with noise. Kdd, 1996) (2013)). In selecting this approach we compared various approaches and we found Meanshift to perform best with respect to identifying clusters closely positioned along a fibre. For example, as shown below, we compared the performance of Meanshift with DBscan in cluster identification with varying radius input parameter; the heatmap scale is in nm.
We concluded that MeanShift provides a more robust cluster identification and that the radius value of 2Xsigma derived from the PCF is appropriate.

• No statistical tests were performed in the study, although no comparisons between experimental conditions are shown. The authors mention the number of replicates performed only on Fig 3 ("13 regions of interest (ROI) were selected from 3 cells"), which is a low number of cells analyzed. The authors should enlarge the analyzed datasets and mention how many
replicates were used throughout the work. As mentioned above, the quantitative analysis originally presented used the data with the constant, highest frame number (3 cells, 895 clusters). In the revised paper we present the analysis of the wider data set (12 cells, 2473 clusters). In practice this makes very little difference to the results, e.g the median cluster FWHM remains at 52 nm.
• The text is well organized and clear as well as the figures but the authors should revise the text in the section "linking RNA polymerase distribution and Y loop transcription" to include missing references to Figure 10. For example, the statement "This distribution of nascent transcript along the Y loops fits with the distribution of RNA-PSer2," refers to results shown in Fig 10A. We are pleased the reviewer found the text and figures well organised and clear. We have added/modified figure references in this section to clarify the relationship between text and figures.
• At end of page 5 the authors describe the location of RNA Pol2 pSer2 with respect to nucleosomes/RNA. As these results are related to findings reported by Castells-Garcia Nucl Acids Res 2022, cited elsewhere in the manuscript, the authors could cite the work here. Done.
• It is unclear for which experiments goat anti-mouse Ig-Alexa Fluor 405 has been used, remove from antibody list if not relevant. This antibody has been removed.
• The use of the 488 laser as the activatory laser is peculiar, can the authors better explain this choice? Empirically we found the 488 laser performed better as an activatory laser-as it gave a higher signal to noise than the more commonly used 405 laser. We have added this statement to the M&M.

Description of analyses that authors prefer not to carry out
Reviewer #2 • It is unclear why the authors did not perform STORM on EU-labeled samples immunolabeled with Pol2 antibodies. In principle, they could have applied the same imaging protocol used in previous figures gaining spatial resolution to better characterize Y nascent transcripts and how Pol2 is arranged with respect to nascent RNA.
There are indeed a number of questions that can be approached by super-resolution imaging of nascent transcripts in association with co-labelling for RPol. We are exploring this interesting