A role for RNA and DNA:RNA hybrids in the modulation of DNA repair by homologous recombination
D’Alessandro et al. proposes mechanistic advances for the role of damage-induced long-non-coding RNAs (dilncRNAs) in double-strand break (DSB) repair. The authors have shown that DNA:RNA hybrids can accumulate at targeted DSBs in transcribed and non-transcribed regions. In addition, damage-induced hybrids accumulate predominantly in S/G2 and are promoted by DSB resection. Although inhibition of transcription did not reduce resection, the authors reported that it reduced recruitment of HR proteins. In perhaps their most compelling result, the authors demonstrate that sequence-specific inhibition of dilncRNAs corresponding to a targeted break result in impaired homologous recombination. Finally, the authors find evidence that RNase H2A is recruited to DSBs in a complex with BRCA2 and interestingly, BRCA2 depletion leads to accumulation of DNA:RNA hybrids at DSBs. These results lead them to suggest a novel mechanism in which DNA:RNA hybrids promote recruitment of HR factors and RNase H2 in the early stages of repair, followed by degradation of the DNA:RNA hybrid by RNaseH2 and repair of the break via HR.
We found the results showing that DR-GFP signal goes down with ASOs around the breaksite to be both surprising and exciting. It was useful to learn that ASOs can be used to specifically inhibit nuclear RNAs which is distinct from inhibiting cytoplasmic RNAs. In addition, the dual observations that BRCA2 recruits RNase H2A and siBRCA2 increases hybrid accumulation provide nice support for the authors’ model regarding removal of hybrids at late stages of repair.
The study led us to many questions and future exciting directions this research could take including:
1. If you only block hybrids on one side of the breaksite, do you get the same reduction in HR?
2. Does chewing the hybrid up result in a different outcome than unwinding? Why does BRCA2 bring RNase H2A instead of bringing Senataxin?
3. Does this mean that BRCA-null tumors have accumulations of DNA:RNA hybrids at DSBs?
4. BRCA1 and BRCA2 have roles in replication as well – do they also recruit RNase H2A to replication forks by a similar mechanism?
For the results in Figure 1, it is important to formally demonstrate that the non-genic regions are not being transcribed to claim de novo recruitment of RNA Pol II. We thought that RNA-seq would be an appropriate control to show this in your system.
We were concerned that the results presented do not fully support the following statement:
“These results show that dilncRNAs, while not affecting DNA-end resection, contribute to the recruitment of HR proteins to the site of damage and, by doing so, they promote HR.”
We felt this statement was an over-interpretation of the data and needs to be softened or supported with more data. The data show that α-amanitin and DRB treatment lead to less recruitment of HR factors and ASOs lead to less DR-GFP signal. α-amanitin and DRB lead to massive transcriptional profile changes in the cell, thus it is difficult to extrapolate from their effects. In addition, α-amanitin and DRB are also two steps removed from the existence of the dilncRNAs at the breaksite. However, the model would be greatly strengthened by experiments that combine site-specific ChIP at the I-Sce1 break with ASOs showing reductions in recruitment of BRCA1, BRCA2, and RAD51.
We felt the italicized text below represented an assertion that was too strong based on the data,
“Based on our observations that dilncRNAs form DNA:RNA hybrids in S/G2-phase cells and control the recruitment of HR proteins to sites of DNA damage, we sought to test the involvement of DNA:RNA hybrids in the focal accumulation of HR proteins at DSBs.”
Focal accumulation of HR proteins was attenuated but clearly still occurs.
The authors need to show that the ASOs are reducing the abundance of the I-Sce1-induced lncRNAs in the experiments of Figure 3e and f; this was an important missing control.
In Figure 4, the stoichiometry of the in vitro binding is important for the proposed mechanism. Also, a negative control is important to assert specificity of binding.
In Figure 6, if “sorting of the S/G2-phase cells population was necessary since BRCA2 inactivation affects the cell cycle (data not shown)” is true for the experiments in 6d, then it must also be true for the experiments in 6a which appeared to be done in asynchronous cells.
Finally, we felt that the overall model would be strengthened by more discussion of the authors’ previous recent paper, Michelini et al., 2017. For example, how do DDRNAs fit into this model? In particular, how does their role differ from the DNA:RNA hybrid?
1. In Fig 1c, it would be more convincing to include a negative control (without I-PpoI) to show the background level.
2. In Fig 2b, the zoomed fields do not match the squared areas in the upper image.
3. In Fig 2c and 4b, we found the y-axis metric, “Normalized number of yH2A.X-DNA-RNA hybrids overlaps relative to random” confusing. It would be clearer the calculation of this metric was included in the figure legend.
4. In Supplementary Figure 3h, G1 and G2 fractions should be different colors.
5. Figure 5a and b – should indicate how far away from the breaksite the primers are and what the ‘unrelated’ site is (either in the figure or legends).
6. It would be less confusing to represent Figure 6c as two different IPs of RNase H2 rather than have two panels of RNase H2A in the same columns.
7. In Fig 6e, annotation for the blue Pac-Man is missing. Is it CtIP? Also, as pictured, the interaction of RPA and DNA:RNA hybrids is confusing (is RPA directly binding the hybrids?) We would consider excluding RPA from the model since there is little evidence for when and how RPA is interacting with the DNA:RNA hybrids.