Differential contribution of steady-state RNA and active transcription in chromatin organization

Abstract

Nuclear RNA and the act of transcription have been implicated in nuclear organization. However, their global contribution to shaping fundamental features of higher-order chromatin organization such as topologically associated domains (TADs) and genomic compartments remains unclear. To investigate these questions, we perform genome-wide chromatin conformation capture (Hi-C) analysis in the presence and absence of RNase before and after crosslinking, or a transcriptional inhibitor. TAD boundaries are largely unaffected by RNase treatment, although a subtle disruption of compartmental interactions is observed. In contrast, transcriptional inhibition leads to weaker TAD boundary scores. Collectively, our findings demonstrate differences in the relative contribution of RNA and transcription to the formation of TAD boundaries detected by the widely used Hi-C methodology.

Keywords: RNA; Hi-C; genomic compartment; topologically associated domains; transcription inhibition.

Figure 1. Overall genome organization is preserved upon RNase digestion

1. Experimental scheme of RNase A digestion of K562 cells before and after formaldehyde (FA) crosslinking (termed as bXL and aXL, respectively) followed by the Hi‐C assay.

2. Agarose gel electrophoresis of RNA extracted from RNase‐treated or control K562 cells.

3. Experimental scheme of ActD treatment of K562 cells followed by Hi‐C analysis.

4. Agarose gel electrophoresis of RNA extracted from control, and 4‐ and 24‐h ActD‐treated K562 cells.

5. Barplots showing qRT–PCR gene expression levels (mean ± SD) of PTEN, FNDC3B, and STAM genes in control, and 4‐ and 24‐h actinomycin D‐treated K562 cells. P‐value: two‐tailed Student's t‐test, (***P < 0.05). The results represent data from three technical replicates from three independent biological preparations.

Figure EV1. Nuclear changes upon RNase treatment in K562 and HeLa cells

1. Immunostaining of active caspase 3 and DAPI staining in RNase‐treated or ActD‐treated and control K562 cells.

2. Immunostaining of Fibrillarin and DAPI staining in RNase‐treated or ActD‐treated and control K562 cells.

3. DAPI staining of HeLa cells before and after crosslinking.

4. Quantification of nuclear area in K562 bXL CTRL and bXL RNase A samples. P‐value: two‐tailed Student's t‐test. The horizontal bands in the boxplots represent the median, the error bars represent ± 1.5 × IQR values, and the outer shapes represent the density of the data points. The sample numbers are indicated below.

Figure EV2. Hi‐C samples show high correlations

1. Dendrogram showing the Hi‐C replicate dissimilarity scores (1‐correlation) of RNase‐treated and control samples based on 1st eigenvalues.

2. Dendrogram showing the Hi‐C replicate dissimilarity scores (1‐correlation) of transcriptionally inhibited and control samples based on 1st eigenvalues.

Figure 2. RNA depletion results in reduction in B‐type compartmental interactions

1. A

   Hi‐C interaction heatmaps at 500‐kb (left) resolution showing all of chromosome 11, and at 100‐kb resolution (right) showing chr11: 79–115 Mb for aXL control and RNase A‐treated cells. The arrows represent similar patterning of long‐range interactions in samples treated with RNase after crosslinking.

2. B

   Hi‐C interaction heatmaps at 500‐kb (left) resolution showing all of chromosome 11, and at 100‐kb resolution (right) showing chr11: 79–115 Mb for bXL control and RNase A‐treated cells. Black arrow indicates the reduction in long‐range interactions in the bXL RNase A dataset. The arrows represent perturbed patterning of long‐range interactions in samples treated with RNase before crosslinking.

3. C, D

   Pearson correlation heatmaps showing the genomic interactions at 500‐kb resolution for chromosome 4, with the 1^st principal components (1^st PC) below for aXL (C) and bXL (D) samples. The bXL RNase A sample shows reduced Pearson correlations across the genome.

4. E

   Violin plot showing the first eigenvalues (calculated by the trans‐chromosomal data) in control and RNase‐treated cells before and after crosslinking. The bXL RNase A sample shows reduction in the negative eigenvalues, indicative of perturbation of B‐type compartments. P‐value: Wilcoxon rank‐sum test. In the violin plots, the horizontal bands represent the median, the error bars represent ± 1.5 × IQR values, and the outer shapes represent the density of the data points. The figures represent data generated from the pooled Hi‐C datasets with two biological replicates.

5. F

   Saddle plots showing the compartmental interactions for both cis‐ and trans‐contacts across the conditions. The bXL RNase A samples display reduced B‐B‐type interactions.

Figure EV3. Compartment switching in RNase‐ and ActD‐treated cells

1. A

   Violin plot showing the genome‐wide Pearson correlation scores of control and RNase‐treated cells, with median values depicted as black dots. The figure is prepared from the pooled Hi‐C data generated with 2 biological replicates.

2. B, C

   Bar plots showing the CTRL to RNase A treatment stable (B) and switched (C) compartments for the aXL (black) and bXL (red) and randomized (gray) samples. Error bars: standard deviation.

3. D

   Bar plots showing the stable and switched compartments for control and 24‐h actinomycin D‐treated cells as well as the randomization averages.

Figure 3. RNA depletion does not affect TAD boundaries

1. A

   Interaction heatmaps at 40‐kb resolution zooming in chr21: 19.5–33 Mb and showing the TAD structures in aXL and bXL datasets. The aXL and bXL RNase A/control log2 ratios are shown below the heatmaps.

2. B

   Boxplot showing the TAD boundary scores for the aXL and bXL datasets. P‐value: Wilcoxon rank‐sum test (***P < 0.05). The horizontal bands in the boxplots represent the median, the error bars represent  ± 1.5 × IQR values, and the outer shapes represent the density of the data points. The figures represent data generated from the pooled Hi‐C datasets with 2 biological replicates.

3. C, D

   Mean interactions frequencies ± 1 Mb centered on all TAD boundaries for each dataset at 40‐kb resolution, as well as the log2 ratios of aXL (left panel) or bXL (right panel) samples. The TAD structures are not altered upon RNase A treatment in samples treated with RNase after crosslinking.

Figure 4. Genomic compartments are largely preserved upon transcriptional inhibition

1. Hi‐C interaction heatmaps at 500‐kb (left) resolution showing all of chromosome 11, and at 100‐kb resolution (right) showing chr11: 79–115 Mb for control and 24‐h ActD‐treated cells. The arrows point out to perturbed long‐range interactions at shorter scales.

2. Pearson correlation heatmaps showing the genomic interactions at 500‐kb resolution for chromosome 4, with the 1^st principal component (1^st PC) below for control and 24‐h ActD‐treated cells.

3. Violin plot showing the first eigenvalues (calculated by the trans‐chromosomal data) in control and ActD‐treated cells. P‐value: Wilcoxon rank‐sum test. In the violin plots, the horizontal bands represent the median, the error bars represent  ± 1.5 × IQR values, and the outer shapes represent the density of the data points. The figures represent data generated from the pooled Hi‐C datasets with 2 biological replicates.

4. Saddle plots showing the compartmental interactions for both cis‐ and trans‐contacts across the conditions.

5. Scaling plot generated by 500‐kb resolution Hi‐C data showing the interaction frequency as a function of genomic distance.

Figure 5. Transcriptional inhibition results in weakening of TAD boundary strength

1. Interaction heatmaps at 40‐kb resolution zooming in chr9: 104‐114 Mb and showing the TAD structures, as well as the log2 ratios, in control and transcriptionally inhibited datasets. Black arrow indicates the increase in inter‐TAD interactions.

2. Boxplot showing the TAD boundary scores in control and 24‐h ActD‐treated cells. P‐value: Wilcoxon rank‐sum test. The horizontal bands in the boxplots represent the median, the error bars represent  ± 1.5 × IQR values, and the outer shapes represent the density of the data points. The figures represent data generated from the pooled Hi‐C datasets with 2 biological replicates.

3. Mean interaction frequencies centered on all TAD boundaries  ± 1 Mb for each dataset at 40‐kb resolution, as well as the log2 ratios. The inter‐TAD interactions are increased upon transcriptional inhibition.

4. Aggregate peak analysis based on the K562 loops coordinates identified in 4. The heatmaps are centered on the loop positions  ± 250 kb using 25‐kb resolution Hi‐C data. There is a ˜50% reduction in loop intensity as assessed by z‐scores and log2 fold change.

5. Boxplot showing the TAD boundary scores in control and 24‐h ActD‐treated cells for boundaries that are bound with differing numbers of CTCF binding sites. The reduction in TAD boundary scores upon ActD treatment is independent of CTCF binding. P‐value: Wilcoxon rank‐sum test. The horizontal bands in the boxplots represent the median, the error bars represent ± 1.5 × IQR values, and the outer shapes represent the density of the data points. The figures represent data generated from the pooled Hi‐C datasets with 2 biological replicates.

6. Scaling plot generated by 40‐kb resolution Hi‐C data showing the interaction frequency as a function of genomic distance, with upper distance limit of 3 Mb. The 24‐h ActD‐treated samples show a slower rate of decay at shorter distances, correlating with the reduction in insulation scores.

Figure 6. Model summarizing the nuclear structure changes observed upon different treatments

RNase treatment of cells before crosslinking results in reduction in long‐range interactions specifically at B‐type compartments, but does not significantly alter TAD boundaries. On the other hand, transcriptional inhibition leads to significant weakening of TAD boundaries without drastic alterations in genomic compartmentalization.