Formation of 53BP1 foci and ATM activation under oxidative stress is facilitated by RNA:DNA hybrids and loss of ATM-53BP1 expression promotes photoreceptor cell survival in mice

Background: Photoreceptors, light-sensing neurons in retina, are central to vision. Photoreceptor cell death (PCD) is observed in most inherited and acquired retinal dystrophies. But the underlying molecular mechanism of PCD is unclear. Photoreceptors are sturdy neurons that survive high oxidative and phototoxic stress, which are known threats to genome stability. Unexpectedly, DNA damage response in mice photoreceptors is compromised; mainly due to loss of crucial DNA repair proteins, ATM and 53BP1. We tried to understand the molecular function of ATM and 53BP1 in response to oxidative stress and how suppression of DNA repair response in mice retina affect photoreceptor cell survival. Methods: We use the state of art cell biology methods and structure-function analysis of mice retina. RNA:DNA hybrids (S9.6 antibody and Hybrid-binding domain of RNaseH1) and DNA repair foci (gH2AX and 53BP1) are quantified by confocal microscopy, in retinal sections and cultured cell lines. Oxidative stress, DNA double strand break, RNaseH1 expression and small-molecule kinase-inhibitors were used to understand the role of ATM and RNA:DNA hybrids in DNA repair. Lastly, retinal structure and function of ATM deficient mice, in Retinal degeneration 1 (Pde6brd1) background, is studied using Immunohistochemistry and Electroretinography. Results: Our work has three novel findings: firstly, both human and mice photoreceptor cells specifically accumulate RNA:DNA hybrids, a structure formed by re-hybridization of nascent RNA with template DNA during transcription. Secondly, RNA:DNA-hybrids promote ataxia-telangiectasia mutated (ATM) activation during oxidative stress and 53BP1-foci formation during downstream DNA repair process. Thirdly, loss of ATM -in murine photoreceptors- protract DNA repair but also promote their survival. Conclusions: We propose that due to high oxidative stress and accumulation of RNA:DNA-hybrids in photoreceptors, expression of ATM is tightly regulated to prevent PCD. Inefficient regulation of ATM expression could be central to PCD and inhibition of ATM-activation could suppress PCD in retinal dystrophy patients.


Introduction
Photoreceptors are light-sensory neurons and one of the six major cell types in the retina, which are organized into stratified layers (Figure 1a). Mutations in more than 250 genes, both retina-specific and ubiquitously expressed, are associated with inherited retinal dystrophies (IRDs) (https://sph.uth.edu/ retnet). These genes have approximately 20 known cellular functions 1 . Mutations in genes coding for proteins involved in retina-specific functions (i.e., phototransduction, visual cycle, retinal development, etc.) could well explain photoreceptor dysfunction or degeneration. But, the reasons that mutations in ubiquitously expressed genes can result in PCD remains unresolved.
A prime example is mutations in ubiquitously expressed members of the U4/U6-U5 tri-snRNP particle (PRPF31, PRPF3, PRPF4, PRPF6, PRPF8) and splice-complex proteins (SNRNP200 and PAP1), which are the second-most frequent cause of autosomal dominant forms of retinitis pigmentosa (adRP) after mutations in rhodopsin [1][2][3] . An exception is DHX38, a spliceosome complex associated RNA-helicase, which has an autosomal recessive pattern of inheritance 4 . Heterozygous mutations in human PRPF genes does not affect any cell type but specifically causes PCD. What makes photoreceptor cells more susceptible to mutations in PRPF is presently unknown. Wheway et al. reported, using mouse cells, that PRPF6, 8, and 31 are important for ciliogenesis 5 . Photoreceptor cells are specialized sensory cilia and defects in ciliogenesis can primarily affect photoreceptor biogeneisis and survival, as seen for mutations in genes such as CEP290 and BBS 1 . However, unlike those in CEP290, mutations in PRPF do not cause PCD in mice. Mice knockout models of PRPF31, PRPF3 and PRPF8, as well as knock-in models containing analogous mutations, do not show any photoreceptor degeneration 6,7 . It was also hypothesized that PRPF-mutations could have more pronounced effect on the splicing of photoreceptor-specific genes, thus specifically deteriorating the health of photoreceptors. As observed, haploinsufficiency of PRPFs causes genome-wide splicing defects and does not explain the photoreceptor-specific phenotype 1,8,9 . Evidently, some other factor or combination of factors causes the higher vulnerability of photoreceptors to the loss of splicing proteins.
Inefficient splicing or defects in mRNP biogenesis can lead to genomic instability by an RNA:DNA-hybrid-dependent mechanism 10,11 . RNA:DNA-hybrids are formed by re-hybridization of nascent RNA with negatively supercoiled DNA behind the moving RNA polymerase, and plausibly accompanied by a single-stranded non-template DNA, to form a three-stranded structure known as R-loop 11,12 . RNA:DNA hybrids are shown to cause DNA breaks in replication-dependent as well as replication-independent manner, mainly by impeding transcription and replication progression 11,13,14 .
Although RNA:DNA hybrids have not been observed in postmitotic neurons; their role in neurodegeneration is alleged 15 . Many proteins involved in RNA:DNA-hybrid dissolution are associated with neurodegeneration. RNA:DNA-hybrid helicases such as senataxin (SETX) and aquarius (AQR), are associated with ataxia with oculomotor apraxia type 2 (AOA2) and type 1 (AOA1) [16][17][18] . Nucleotide excision repair proteins, ataxia telangiectasia mutated (ATM) and Fanconi anaemia pathway proteins, which are associated with neurodegeneration, neurodevelopmental defects or microcephaly, have recently been implicated in RNA:DNA-hybrid resolution 13,14,19,20 . This led us to speculate that RNA:DNA hybrids could be formed in retinal neurons and could play a role in retinal degeneration.

RNA:DNA hybrids specifically accumulate in photoreceptor cells of retina
We checked if post-mitotic retinal neurons could accumulate RNA:DNA hybrids. Mice retina was stained with S9.6 (RNA: DNA-hybrid-specific) antibody. Higher levels of RNA:DNAhybrids are observed in adult photoreceptor nuclei than in the other retinal neurons (Figure 1b). Nuclei of murine photoreceptors have an inverted chromatin organization, with central heterochromatin and peripheral euchromatin (Supplementary Figure 1a). Interestingly, the RNA:DNA-hybrids are observed on the peripheral euchromatin region of photoreceptor nuclei, in proximity to the nuclear membrane (Figure 1b, middle panel). We also checked the expression of RNAseH1 (a ribonuclease) and Senataxin (a helicase), which are enzymes involved in dissolution of RNA:DNA hybrids. Senataxin is expressed mainly in the outer nuclear layer (ONL) of the retina and localized to the euchromatin area of photoreceptor nuclei (Supplementary Figure 1b). Contrastingly, RNaseH1 is mainly expressed in the ganglion cell layer (GC) and inner nuclear layer (INL) of the retina, but not in the ONL (Figure 1a and Supplementary Figure 1c). Extra-nuclear staining of RNAseH1 (likely mitochondrial) in the photoreceptor inner segment is observed.
Loss of PRPF31 cause genomic instability but not in retinal neurons As photoreceptors can accumulate RNA:DNA hybrids, we wondered whether loss of spliceosomal proteins could lead to RNA:DNA-hybrid-dependent genomic instability in photoreceptor cells. Of the eight PRPFs, PRPF31 gene aberrations are a major cause of adRP (i.e. RP11) and lead to genome-wide splicing defects 8,9 .
We used siRNA-based PRPF31 downregulation in RPE-1 cells and quantified foci formation of early DNA damage and repair markers, i.e. γH2AX (H2AX phosphorylated at ser139) and 53BP1. A significant increase in γH2AXand 53BP1 foci is observed in cells depleted of PRPF31 (Figure 2a, b and Supplementary Figure 2a). We next analyzed primary cells from the stromal vascular fraction (SVF) of PRPF31-deficient mouse models (Prpf31 +/and Prpf31 +/A216P ) 6,7 . The Prpf31-A216P variant reduces the stability and nuclear localization of U4/U6-U5 tri-snRNP complex 21 . Primary SVF cells obtained from heterozygous Prpf31 +/A216P mice show clear accumulation of γH2AX (Figure 2c, d). Notably, expression of active RNaseH1 in these cells significantly reduced both γH2AX and 53BP1 signal. This indicates the role of RNA:DNA hybrids in genomic instability observed in the absence of functional PRPF31. Cells obtained from Prpf31 +/mice also exhibit accumulation of γH2AX (Supplementary Figure 2b). We next assessed whether PRPF31-deficient photoreceptors also show increased genomic instability. But unlike RPE-1 and primary SVF cells; no elevation in genomic instability was observed in the retinal neurons of adult Prpf31 +/ A216P mice (data not shown). In the retina of postnatal day 20 mice, an increase in γH2AX and 53BP1 foci was observed ( Figure 2e). Notably, 53BP1 is not expressed in the ONL (composed of photoreceptor nuclei), except in the apical (outermost) layer of photoreceptors nuclei (Figure 2e, arrow).

Photoreceptor cells show slower DNA repair, independent of ATM and 53BP1
The fact that adult mouse photoreceptors can accumulate RNA: DNA hybrids, but do not show any accumulation of genomic instability markers is puzzling. To understand why this is the case, we looked at DNA repair markers in irradiated photoreceptor cells. As reported previously 22 , we also observed that mouse photoreceptor cells have inefficient DNA repair. Irradiation induces γH2AX formation in all retinal cell types, but localizes only to the euchromatin region (Figure 3a, b). As aforementioned, 53BP1 is not observed in ONL (containing the nuclei of photoreceptors) and the outer half of the INL (composed mainly of horizontal cell nuclei) (Figure 1a, Figure 3a). Only at 24 h post irradiation did the γH2AX signal disappear from the nuclei of photoreceptors (Figure 3c). We also checked for irradiationinduced cell death by terminal deoxynucleotidyl transferase dUTP nick-end labeling (TUNEL) assay. The retinal neurons show resistance to irradiation-induced cell death and, unlike the ganglion cell layer, ONL showed no TUNEL-positive cells until 24 h post-irradiation (Supplementary Figure 3).
ATM is a major PI-3 kinase for post mitotic neurons. We observed that in photoreceptors, most of irradiation-induced H2AX phosphorylation is independent of ATM 22 . Absence of functional ATM or the presence of ATM inhibitor does not inhibit H2AX phosphorylation in the ONL of the retina (Figure 4a, b). Consistently, western blot analysis of microdissected neural retinas showed that 53BP1 and ATM levels were depleted around post-natal day 20 (Figure 4c). Although analysis of cDNA from the neural retina shows that an alternative spliced form of ATM, lacking the N-terminal PRD and C-terminal FATC domains, could be present (Figure 4d).
RNA:DNA hybrids effectuate ATM-activation in presence of oxidative stress Inefficient DNA repair and an absence of ATM is not expected in photoreceptors, especially considering that markers of oxidative stress are most pronounced in the cerebellum and retina 1,23,24 . Oxidative DNA damage is the most common cause of DNA damage in post-mitotic neurons and could result in single-strand breaks, a major cause of neurodegeneration. Antioxidant treatments have proved to be promising neuroprotective strategies for many retinal dystrophies 25,26 .
ATM is a sensor of oxidative stress 27,28 . DNA topology or subtle chromatin changes could also activate ATM, even in absence of a DNA break 29 . Activated ATM can signal DNA repair, cell cycle arrest and also cell death mainly by a p53 dependent pathway 30,31 . Noticeably, ATM is shown to promote RNA:DNA hybrid formation on transcribed sites by removal of spliceosomal complex 32 . We wondered if an absence of ATM is linked to the presence of RNA:DNA hybrids and high oxidative stress in mice photoreceptors. To assess whether RNA:DNA hybrids can directly affect ATM function during oxidative stress; we looked at H 2 O 2 induced ATM activation in presence and absence of ectopically expressed RNAseH1. Notably, removal of RNA:DNA hybrids by RNaseH1 overexpression completely abolishes ATM phosphorylation at Ser1981 after H 2 O 2 treatment, as observed by western blotting and immunofluorescence (Figure 5a, b). The suppression was stronger than that obtained using an ATMspecific inhibitor. We also observed that the H 2 O 2 -induced activation of ATM was more prominent in proximity to nuclear membrane ( Figure 5b).
As ATM activation depends on chromatin association and release, we quantified nuclear ATM in detergent-permeabilized cells expressing inactive hybrid-binding (HB) domain (which stabilizes RNA:DNA hybrids) or active RNaseH1 (which destabilizes the hybrids) 13 . The results show high variability, with possibly multiple factors controlling the association of ATM with chromatin; however, it seems that stabilization of hybrids could increase the nuclear retention of ATM (Supplementary Figure 4a). The amount of ATM cross-linked to insoluble pelleted chromatin increased after H 2 O 2 treatment and could be partially released by over-expressing RNAseH1 (Supplementary Figure 4b). This suggests that RNA:DNA hybrids regulate the interaction of ATM with chromatin.
RNA:DNA hybrids co-localize and promote 53BP1 foci formation We next looked at 53BP1, which is also absent from photoreceptors. 53BP1 is a pro-non-homologous end-joining protein and accumulates on damaged DNA sites in an ATM-dependent manner 33,34 . The formation of 53BP1 foci is crucial for DNA repair, checkpoint activation and cell death 33,35,36 . Notably, stabilization of RNA:DNA hybrids via the expression of the HB domain significantly increases 53BP1 foci formation. Conversely, RNAseH1 overexpression significantly suppresses 53BP1 foci formation ( Figure 6a).
Notably, high-resolution images using confocal microscopy showed that 53BP1 foci consistently co-localize with RNA:DNA hybrids (Figure 6b RNA:DNA hybrids promote ATM-mediated DNA repair Clearly, DNA repair activity of ATM and 53BP1 depend on RNA: DNA hybrids. We next assessed whether ATM or 53BP1 are crucial for RNA:DNA-hybrid removal. Using the HB domain, we probed and quantified RNA:DNA-hybrids in cells depleted of ATM. Notably, the number of RNA:DNA hybrids decrease in cells depleted of ATM (Figure 7a, b). As shown before, ATM promotes RNA:DNA hybrid accumulation 32 . This action require ATM phosphorylation, as ATM inhibitor treatment suppresses RNA:DNA hybrid formation ( Figure 7c). Expectedly, ATM-depleted cells are also inefficient in forming 53BP1 foci ( Figure 7a). The results supports the idea that the ATMdependent 53BP1 association with chromatin and RNA:DNA hybrid formation are interdependent events. RNA:DNA hybrids are primarily considered to be a source of genomic instability. Notably, we observe higher levels of oxidative stress-dependent genomic instability when RNA:DNA hybrids are removed. In Figure 7d, as expected, stabilization of RNA:DNA hybrids by the HB domain could increase γH2AX foci levels. However, when treated with H 2 O 2 , HB-domain expressing cells do not show further increases in γH2AX foci. Contrastingly, destabilization of RNA:DNA hybrids by RNaseH1 overexpression results in manifold γH2AX foci accumulation after H 2 O 2 treatment. Similar results are obtained in western blot analysis, wherein the cells over-expressing RNAseH1 show increase in γH2AX after H 2 O 2 treatment ( Figure 6a). Very likely, RNA:DNA hybrid formation is a crucial step during ATM-mediated DNA repair, and higher levels of γH2AX in the absence of ATM-activation is a result of prolonged DNA repair and damage accumulation.
Absence of ATM slow down photoreceptor cell death in rd1 mice ATM is crucial for repair as well as induction of apoptosis. In the presence of unrepaired DNA breaks or blocked DNA-protein complex intermediates, ATM can initiate signaling of the cell death pathway 31 . ATM-deficient cells are defective in irradiationinduced apoptosis 30 . As aforementioned, the ONL of the mouse retina shows resistance to irradiation-induced cell death, as observed by TUNEL staining (Supplementary Figure 3). ATMinduced cell death is via p53 dependent pathway 31 . It is known that ectopic expression of p53 in photoreceptor cells promotes photoreceptor cell death 37 , although the mechanism is unclear 38 .
We observed that stabilization of hybrids by the HB domain in RPE-1 cells leads to S-phase accumulation possibly by RNA: DNA-hybrid-dependent inhibition of replication fork progression (Supplementary Figure 5) 10,11 . Notably, unlike RPE-1 cells, HBdomain expression in HT1180 cells shows an accumulation of sub-G1 apoptotic cells, but no defects in S or G2 phase progression (Supplementary Figure 5). The different outcomes show that stable RNA:DNA hybrids not only create replication stress but can also induce cell death, possibly via the apoptotic pathway.
Photoreceptors are post-mitotic terminally differentiated neurons, thus RNA:DNA-hybrid-induced replication stress cannot occur. However, in the presence of high oxidative stress in photoreceptors, RNA:DNA hybrid dependent constitutive ATM activation could promote cell death 30,31 . We thus wondered whether the absence of ATM in retinal neurons promotes photoreceptor cell survival during high oxidative stress and metabolic demands.
As ATM is only expressed before postnatal day 20 in the mouse retina ( Figure 4c, d), we used the rd1 mouse model, which has a mutation in Phosphodiesterase 6B gene and loses 80 percent of photoreceptors before postnatal day 15 due to severe oxidative stress 39 . We removed ATM in the Pde6b -/background and analyzed the retina of the animal at p20. In the retina of ATM knockout mice (Pde6b -/-Atm -/-), photoreceptor cell death decelerates and the thickness of the ONL is significantly higher (Figure 8a, b).
We next resorted to electroretinogram (ERG)-based functional analysis of the retina, which measures the light-induced transretinal flux of ions. ERG of Pde6b -/-Atm -/mice shows a slight but significantly improved b-wave in dark-adapted animals, indicating the protection of retinal function, as compared to Pde6b -/-Atm +/+ animals ( Figure 8c).
We next assessed what happens in human photoreceptors. As observed by S9.6 antibody staining, RNA:DNA hybrids are  also observed in human photoreceptor nuclei, though unlike mice photoreceptors they are in central euchromatin region (Supplementary Figure 6a). Notably, unlike the murine photoreceptors, expression of ATM and 53BP1 in adult human retina is observed by immunofluorescence (Supplementary Figure 6b). All data are available on OSF 40 .

Discussion
RNA:DNA-hybrid-dependent ATM hyper-activation during oxidative stress could promote photoreceptor cell death. The mechanism could be more pronounced in human patients but not murine RP models, as former express higher levels of ATM and 53BP1. This could possibly explain why some mouse models, e.g. Prpf31 mutants, which cause retinitis pigmentosa-11 in humans, do not exhibit photoreceptor cell death 1,7 . Detailed studies are required, but we anticipate that expression of ATM and 53BP1 are fine-tuned in photoreceptors with respect to RNA:DNA-hybrid and oxidative stress levels (Supplementary Figure 7). Loss of this equilibrium by increased RNA:DNA-hybrid accumulation or higher expression of ATM could cause PCD in progressive retinitis pigmentosa and age-related macular degeneration 1,4 . Loss of ATM and 53BP1 in mice photoreceptors is a compromise in which decreased sensing for DNA damage is coupled to slower activation of cell death signaling in post-mitotic neurons (Supplementary Figure 7). Possibly, unlike mice, humans are diurnal and live much longer, thus would require better DNA repair, and have tighter uncoupling of gene-expression and DNA repair. It was recently reported that osteoclast cells also show better survival in absence of ATM 41 . Though there must be additional factors that influence photoreceptor cell survival, we think our work will help to improve understanding of the mechanism underlying photoreceptor cell death, and propose ATM and 53BP1 as targets of neuroprotection strategies in the human retina.
The molecular mechanisms that sense and resolve RNA:DNA hybrids are unclear and under intense investigation 5,6 . It is known that RNA:DNA hybrids are intimately linked to genomic instability and replication stress. Our work shows that formation of RNA:DNA hybrids is central to the ATM-53BP1-dependent repair pathway. Thus, RNA:DNA hybrids could affect the efficiency of DNA repair, cell death signaling and checkpoint activation (Supplementary Figure 7). We expect that future work to understand RNA:DNA-hybrid-associated molecular pathways will further elucidate its role in neurodegeneration and ageing.

Methods
Cells and cell culture Primary vascular stromal fractions (VGF) were obtained by standard protocol. Confluent cultures of primary VGF cells, RPE-1, HCT1180 and HEK293T cells were all maintained in DMEM (#32430027, Gibco™) supplemented with 10% FBS, at 37°C and 5% CO 2 . To induce oxidative stress, cells were incubated in DMEM containing freshly diluted 500 µM H 2 O 2, for 1 h at 37°C and 5% CO 2 , before being processed for western blot or immunofluorescence.
3Xflag tagged HB-domain AAV viral DNA constructs and RNA:DNA hybrid probing 3X flag tagged HB domain of RNaseH1 and 3Xflag tagged full RNaseH1 lacking the mitochondrial targeting sequence 13 were cloned into a pAAV-IRES-hrGFP (Agilent; #240075) plasmid and viral particles were produced using the AAV Helper-Free System (Agilent #240071), as per the manufacturer's instructions. Cells were transduced in Biosafety level 2 facility and incubated for at least 24 h to allow expression of the tagged protein.

Immunofluorescence of the mouse retinal sections
For immunofluorescence of the mouse retina, animals were euthanized by cervical dislocation and eyes excised and fixed in 4% paraformaldehyde in PBS for 30 min at room temperature. After repeated PBS washes, fixed eyes were incubated at 4°C for 8 h each in 10% sucrose-PBS, 20% sucrose-PBS and 30% sucrose-PBS followed by 50-50 solution of 30% sucrose and optimal cutting temperature (OCT) compound (Tissue-Tex,#4583). The eyes were frozen in 100% OCT in dry ice and stored at -20°C. For cryotome sectioning, performed at -20°C, serial sections 18-mm thick were mounted in five parallel series and stained as described for cells above. For RNA:DNA-hybrid detection using S9.6 antibody, sections were pretreated with Accutase (Sigma; #A6964) for 30 min at room temperature for tissue dissociation and staining was performed as reported by Bhatia et al. 13 . For TUNEL assay the In Situ Cell Death Detection Kit (Roche, Mannheim, Germany), was used as per manufacturer's instructions.

Quantitative analysis of immunofluorescence images
Metamorph (Molecular Devices, Version 7.1) image analysis software was used to quantitate foci, signal intensity and signal area with inbuilt functions, i.e. Granularity and Line Scan. In brief, for foci analysis, maximum projection of z-stacks is created (with LAS AF Leica software) and saved as .tiff images. The images are opened on Metamorph and converted to 8-bit format by inbuilt "Multiply-function". Background subtraction is done based on unstained area in the image. A mask is created, based on DAPI image, to assign the nuclear area. Using the Granularity function, the signal (with minimum granule size in pixels) is quantified per nuclei and updated in a linked excel sheet. To measure ONL thickness in mice retina using metamorph, recoverin-stained retinal section images were opened and three lines per image were drawn perpendicular to the ONL (i.e. recoverin-stained ONL) using the LineScan-tool. The intensity measurements are automatically documented in a excel file. Length of line with pixels containing positive value (for recoverin signal) was used to measure the thickness of ONL. The length in pixels was converted in uM by multiplying with the conversion factor, obtained by measurement of the scale-bar on the image. These measurements were later used for quantitative analysis in GraphPad Prism 5 package software.
Irradiation induced DNA damage in the retina Mice were anesthetized by sub-cutaneous injection of ketamine/xylazine (80/12; mg/kg body weight) and exposed to 5 Gy of gamma irradiation (BioBeam-8000: Gamma Service Medical GmbH). After irradiation, mice were returned to their cages for 1 h (unless specified; as in Figure 3c). Thereafter mice were euthanized and the eyes were processed as mentioned before. To pre-treat the retina with small molecule kinase inhibitors, the cornea is dissected out as described by Donovan et al. 42 . Explants were incubated for 1 h in DMEM+10% FBS medium with 10 µM ATMi (KU55933; ATM inhibitor from Tocris) or 5 µM of DDRi (PI3 kinase inhibitor; a gift from Prof. Oscar Fernandez Capatillo, CNIO, Madrid). Explants were then exposed to 5 Gy of gamma irradiation and incubated in DMEM (with ATMi or DDRi) for 1 h at 37°C and 5% CO 2 . Thereafter, the retinal explant were processed as aforementioned for mouse eyes.

Primers and PCR for ATM expression in neural retina
The neural retina of two mice for each time point (at postnatal days 10, 20, 60 and 270) was microdissected; 1 eye each was used for western blot and one eye was used to prepare total RNA. RNA was quantified using Nanodrop ND-100 Spectrophotometer (Manufactured by NanoDrop Technologies), normalized and reverse transcribed to cDNA using QuantiTect Reverse Transcription Kit (Qiagen,#205311). Primers were designed using the Primer3Plus online tool, on the selected exons of ATM cDNA sequence (from the ENSEMBL database). Primers target a region on ATM mRNA, relative to domains on protein sequence. I.e. betwen exon 2-4 for TAN domain; between exon 62-64 for FATC domain and between exon 13-17 for control FAT domain. The primers (sequence given below) were validated using UCSC in silico PCR, and selected if they produce single PCR product from mouse transcriptome sequence and no product from mouse genomic sequence. PCR reaction was performed using MyTaq TM red DNA polymerase, using standard PCR condition with 54°C of annealing temperature (all primers have have Tm between 58°C and 60°C); for 30 sec and extension at 72°C for 1min, for 30 cycles, for all primers.

Animal handling
All experiments are performed in compliance with Spanish and European Union laws on animal care in experimentation and approved by the Committee of Animal Experimentation, CABIMER, Seville, Spain. Mice are maintained in Specific Pathogen Free (SPF) conditions and health status is monitored through a comprehensive surveillance program. Cages (4-6 adults per cage), bedding (sawdust) and water (sterilized by autoclaving) and food (irradiated Rodent VRF1) were changed weekly (every Tuesday). Room temperature (21°C) and 12-12 light cycle (6 pm -6 am) were maintained. Equipment and material that need to enter the SPF zone were decontaminated by hydrogen peroxide vapor. The number of mice used in the study was kept to a minimum and sample size calculations were not performed prior to experiments due to lack of equivalent datasets and information about expected results. Power analysis was performed retrospectively to confirm that the power of statistical analysis is >0.8, at alpha=0.05.
Retinal sections for DNA response analysis are from C57BL/6J, adult wild-type mice (8-10 weeks) both male and female (22±3 g). S.S. Bhattacharya lab at University College London, previously reported PRPF31 mouse models 6, in collaboration with Charles River (France). Mice were procured and transferred to the SPF animal facility of CABIMER, Seville, Spain. Prpf31 A216P/+ mice were on a mixed background of 129S2/Sv (source of stem cell used for mutation incorporation) and C57BL/6J (background used from crossing the chimeric mice). Prpf31 ± knockout mouse has BALB/c, 129S2/Sv and C58BL/6J mixed background; as they were generated by crossing the Prpf31 A216P/ + mouse with a BALB/c mouse expressing Cre recombinase (BALB/c-Tg(CMV-Cre)1Cgn/J) 6 . Retinal sections were obtained from 20-day old Prpf31 A216P/+ mice. VSF cells were isolated from adult PRPF31 mouse models (8-12 week). ERG and retinal thickness analysis were done on 20-day old Pde6b -/-ATM +/knockout mice. Pde6b -/-ATM +/were produced by crossing ATM +/knockout mouse (originally created in a mixed background (129/SvEv and NIH Black Swiss) 43 with Pde6b -/mice (FVB/Ncrl, from Charles River, an early onset retinal degeneration strain homozygous for allele Pde6brd1). ATM mice were obtained from Felipe Cortes lab (CABIMER, Seville, Spain) and genotyped as reported before 43 To study DNA repair response in irradiated mice, for two independent repeats are performed. For each repeat all animals were used from a single litter of C57BL/6J adult wild-type mice of same-sex (all male or all female).
Weaning was performed for all strains at postnatal day 20. For PRP31 mouse models and Pde6b -/-ATM +/mice, which were studied at postnatal day 20 (15±2 g), pups were always with mother. For ERG analysis, 24 h dark adaptation was performed with mother. For Prpf31 A216P/ + mice retinal sections, two mice each i.e. wildtype and heterozygous A216P/ + (i.e. four total mice i.e. 8 eyes) are analyzed. For VS fraction cells, two mice for each mice model respective wild-type were used i.e. 8 eyes. Independent cultures were maintained, for up to 10 passages.
For ERG analysis, total 24 mice i.e. 48 eyes (from nine ATM +/+ , six ATM +/and nine ATM -/mice) were used, all coming from three litters. Males and females were not distinguished. Animals were genotyped after ERG analysis was performed to reduce any bias. For morphological preservation, retinal thickness analysis was performed for two Pde6b -/-ATM +/+ and three Pde6b -/-ATM -/were used (total five mice). Metamorph software-based automated Linescan analysis of recoverin staining was performed for retinal thickness analysis, described above.
The primary result of our study is that molecular function of ATM and 53BP1 depend on RNA:DNA hybrids; and removal of RNA:DNA hybrids completely inhibit ATM activation during oxidative stress. The primary outcome from mice models is the partial preservation of retinal structure and function after ATM-removal in rd1 mice. The additional outcomes are: suppressed DNA repair response and loss of ATM and 53BP1 expression in retinal neurons, comparison with human photoreceptors and presence of RNA:DNA hybrids close to the proximity of nuclear membrane of murine photoreceptor cells. All efforts were made to ameliorate the suffering of animals.

Electroretinography
Electroretinography was performed using a Color Dome Ganzfeld (Diagnosys LCC, MA, USA) as detailed before by Lourdes et al. 44 . Briefly, to evaluate scotopic vision, mice were dark-adapted overnight and anaesthetized by sub-cutaneous injection of ketamine/xylazine (80/12; mg/kg body weight). A drop each of 10% phenylephrine and 1% tropicamide were used to dilate the pupils of the animal. To detect retinal response, the mouse was placed inside the ColorDome Ganzfeld (Diagnosys LCC, MA, USA) and electrodes were touched on the surface of the corneas, pre-treated with a hydrating agent (1% methylcellulose). A single pulse of white-flash (6500 K) was used for stimulation, with the stimulus strength of 0.1, 1 and 10 lux. An average of 15 responses was made, with an inter-stimulus interval of 15 s.

Statistical analysis
The indicated statistical tests were performed using GraphPad Prism 5 package. In brief, the Kolmogorov-Smirnov normality test was used to check distribution of data (alpha=0.05).
A parametric two-tailed Student's t-test was used for data with normal distribution; otherwise, a non-parametric Mann-Whitney test was applied. For multiple comparisons, one-way-ANOVA followed by Turkey test was used if data had normal distribution; otherwise a Kruskal-Wallis test followed by Dunn's post hoc test was used. Statistical significance is marked by one, two or three asterisks, indicating P < 0.05, P < 0.01 or P < 0.001, respectively.

Data availability
All data associated with this study, including all raw microscopy images and uncropped western blots, are available on OSF: http://doi.org/10.17605/OSF.IO/X3CM7 40 . Data are available under the terms of the Creative Commons Attribution 4.0 International license (CC-BY 4.0).

Competing interests
No competing interests were disclosed.

Grant information
The work is supported by Junta de Andalucia, Spain (P09CT54967) and Juan de la Cierva grants (IJCI-2014-22549) from the Ministry of Economy, Industry and Competitiveness, Government of Spain.
The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. the presence or absence of RNaseH1 were exposed to formaldehyde and lysed in denaturing conditions, and the lysate was analyzed by SDS-PAGE. Loss of signal due to chromatin crosslinking is observed by western blot analysis. The cytoplasmic protein β-actin was used as a control. The representative image of n=4 independent repeats is shown. Click here to access the data. 1.

2.
3. In this manuscript Bhatia et al, provide evidence that suppression of the ATM signaling pathway in murine photoreceptors (PR) is a protective mechanism from oxidative stress/RNA:DNA-hybrid induced apoptosis.

Open Peer Review
According to their proposed model, the high levels of oxidative stress PR are naturally subjected to, promote formation of RNA:DNA hybrids which activate ATM and result in p53-mediated apoptosis. Suppression of ATM could therefore have protective effects against retinal degeneration. They support this model by presenting evidence showing that: 1. PRs have higher levels of RNA:DNA hybrids relative to other cells in the retina. 2. Impaired splicing, that has been linked to R-loop formation in other models, can cause RNA:DNA dependent double-strand breaks (DSBs) in retina-derived cell lines, but not in neurons, which they attribute to lack of ATM and 53BP1 expression by the later. 3. In response to ionizing radiation, despite their lower DSB repair capacity, which is independent of ATM, PRs are resistant to apoptosis 4.Oxidative stress in cells activates ATM through an RNA:DNA hybrid dependent mechanism, 5. Absence of ATM delays cell death in a mouse model of retinal degeneration.
Overall the authors present a large number of interesting data from experiments with mouse retinas and established cell lines (RPE and SVF cells). Their conclusions are consistent with the literature but they also report novel and even controversial findings. The data are well presented and discussed although in some instances a better representation and/or additional controls would strengthen the authors' arguments. On the somewhat negative side, interesting as the data may be, the large number of issues addressed and experimental models used, result in confusion, loss of coherence and some issues are not addressed or discussed sufficiently. Finally, there are a few statements that should be phrased more carefully as they can be unintentionally misleading.
Some issues that could be addressed include: The figure supporting the important conclusion that "RNA:DNA hybrids specifically accumulate in should be presented in a manner similar to that in Supplemental photoreceptor cells of retina" figure 1. Can the authors suggest an explanation for the increased RNA:DNA hybrid levels in PRs? Clearly they cannot be formed as a consequence of oxidative stress-activated ATM, as they see a depletion of ATM from ONL. The authors propose that splicing defects (such as silencing or inactivating mutations of PRPF31) induce RNA:DNA-dependent DSBs in cells lines (RPE-1/SVF) but not in retinal neurons. In cultured cells, increased levels of DSBs (assayed here as gH2AX and 53BP1 foci) could be the result of R-loop dependent replication stress and fork collapse. The difference in DSB levels S-phase and G2-phase cells from the analysis. The same is true for the primary cells from Prpf31 mice. Without such a G1-specific analysis it is not useful to compare the data from a proliferating cell line with strictly postmitotic retinal neurons of adult mice. Prpf31 Another striking finding is the discrepancy between the numbers of gH2AX and 53BP1 foci measured in Fig. 2a and 2b, since many papers have described that the majority of DNA DSBs show both of these markers. The authors should comment on this.
Finally the authors describe an increased number of gH2AX and 53BP1 foci in PRs of 20 day old Prpf31 mice when compared with wild type PRs. This is wrong for several reasons: First, the authors present measurements of signal intensities and not the quantification of foci. This should be changed in the text. Second, there is only a significant increase in 53BP1 signal intensity but not in gH2AX. The authors also state that 53BP1 is only expressed in the apical layer of photoreceptor cells. 3. The authors show that RNAse H1 overexpression has a strong impact on DSB levels. One might wish to see the impact of this overexpression on the level of RNA:DNA hybrids. The authors should present a quantification similar to the one presented in Fig. 7a. 4. The major message of this manuscript is that the loss of ATM-53BP1 expression promote PR survival. As an evidence for ATM loss the authors present data from western blot analysis of micro-dissected neural retinas in Fig.4. I have several problems with these findings and conclusions: First, on page 11 the authors refer to Donovan et al for the procedure of microdissection. After checking this paper one would conclude that the authors have taken the whole retina for this analysis. But in this case these results would not match with the finding that 53BP1 is still present in adult retina (as shown in Fig. 2). Thus, I conclude that microdissection means the isolation of the ONL for western blotting. The authors should clarify their methods. Second, the conclusion that ATM activity is lost seems wrong for the following reason: As Frohns et al have shown by the irradiation of Scid mice and additional experiments, in vitro cone and rod PRs show a residual ATM activity that is able to phosphorylate gH2AX after DSB induction. Thus, ATM activity should be present in PRs rendering the major message of this manuscript questionable.
5. The sequence of data presentation in Fig. 5 and 7 is confusing, making it hard to get a clear model of the role of hybrids, ATM expression and its activation in the presence and absence of DNA damage. The authors should try to create a scheme that helps the reader to understand their data. Furthermore, Fig. 5a indicates decreasing gH2AX levels in H2O2 treated and RNAse H1 overexpressing cells when compared to RNAse H1 overexpression alone. In Fig. 7d the opposite can be observed. Why is that? 6. In Fig. 6a the authors show that H2O2 treatment alone has no impact on the numbers of 53BP1 foci. This is irritating since the same treatment increases gH2AX signals as measured in WB presented in Fig.  5a. Considering the fact that both, gH2AX and 53BP1 are DSB markers the authors should comment on this contradictory finding. In Fig. 6b the authors state that "53BP1 foci consistently co-localize with RNA:DNA hybrids". From 10 53BP1 foci shown in one picture i would score only 5 of them to clearly colocalize with hybrid signals. Thus, the authors should specify that statement or show a quantification.
1 2 colocalize with hybrid signals. Thus, the authors should specify that statement or show a quantification. 7. In Fig. 7d the authors present increasing numbers of gH2AX foci in cells expressing the HB domain. But again the gH2AX analysis is not carried out in a cell cycle specific manner. This is even more of a problem than for Fig. 2 since RPE-1 cells (which likely were used for this analysis, although it is not clearly stated in the legend) show an S-phase accumulation after HB domain overexpression (as shown in Fig.  S5). Thus, the increase in gH2AX foci levels might be due to a higher number of S-phase cells (which usually show higher spontaneous DSB levels) in comparison to WT cells.
8. On page 6 the authors state "Expectedly, ATM-depleted cells are also inefficient in forming 53BP1 foci ( )". I agree that this is expected but still the pictures presented are of too low resolution to Figure 7a underline that statement (no foci visible). Thus, high resolution pictures should be presented. 9. In Fig. S6b it is shown that ATM and 53BP1 expression is detectable in human PRs. Hereby, the fact that ATM staining shows a punctual or foci-like staining is irritating. If this would be due to its accumulation at DSBs, why is there no colocalization with the also clearly visible 53BP1 foci (have the authors checked for pATM staining)?
Minor issues: On page 4 the authors say that the outer half of the INL is composed mainly of horizontal cell nuclei which is not correct. Horizontal cells are a minority in the outer INL. The most common cell type in this region are bipolar cells. The authors should correct that.

If applicable, is the statistical analysis and its interpretation appropriate? Yes
Are all the source data underlying the results available to ensure full reproducibility? Partly The manuscript seeks to address the underlying mechanisms of photoreceptor degeneration, particularly that due to ubiquitously expressed proteins. They hypothesize that R-loops may play a role, as several splicing and RNA helicase proteins have been implicated in degeneration and R-loop formation. Using the S9.6 antibody, they find that R-loops are formed in photoreceptor cells of the ONL at higher levels than other cell types. They reason that splicing defects could further increase DNA damage in this high R-loop environment and deplete PRPF31, mutations in which are associated with photoreceptor degeneration. In both RPE1 cells depleted for PRPF31 and primary SVF cells from mice with a PRPF31 knockin mutation, increased levels of gH2AX and 53BP1, markers of DNA breaks, are observed. In the Ki cells, this is suppressed by expression of RNAseH1 that will degrade R-loops, supporting the proposition that the damage is R-loop dependent. However, this signaling was not observed in the ONL layer in vivo but it appears that 53BP1 expression is very low compared to other tissues. They next demonstrate that in response to IR, DNA repair assessed by the appearance and loss of gH2AX is slower and independent of ATM and 53BP1, both of which appear to be lowly expressed in the mature retina. Interestingly, they identify a truncated form of ATM mRNA expressed in the retinal cells that may lead to a null or kinase dead ATM protein. Next they show that peroxide activates ATM in a manner dependent on RNAseH1 (thus, presumably R-loops) but in contrast, 53BP1 foci, that co-localize with R-loops are suppressed by peroxide. Analysis of ATM deficient cells using the HB domain to stabilize R-loops revealed fewer R-loops and 53BP1 foci, suggesting that ATM promotes R-loops formation to recruit 53BP1. Unexpectedly, gH2AX signal is strongly increased in the presence of peroxide and RNAseH1 expression, potentially reflecting deficient repair. Finally, the authors show that ATM loss promotes photoreceptor survival and ONL thickness in a mouse model.
The manuscript is overall well written and the data clearly presented. There are a number of new and interesting findings regarding the expression of damage factors and their influence on the retina. However, the logic and flow of the manuscript are somewhat confusing in places. The authors switch between post-mitotic tissues and proliferating cells and there is some apparently conflicting data and overstatements that could use clarification. Here are some specific comments and suggestions to address these points.
Under the header "RNA:DNA hybrids specifically accumulate in photoreceptor cells of retina" the authors state: "Higher levels of RNA:DNAhybrids are observed in adult photoreceptor nuclei than in the other retinal neurons". Based on the data provided, this claim, that is central to the manuscript, is not well substantiated. Essentially a single cell representing multiple cell types is shown in Figure 1b in comparison to the ONL cells. To support this definitive statement that , a quantitative or even semi-quantitative analysis (for R-loops are specifically higher in the ONL

If applicable, is the statistical analysis and its interpretation appropriate? Yes
Are all the source data underlying the results available to ensure full reproducibility? Yes Are the conclusions drawn adequately supported by the results? Partly No competing interests were disclosed.

Competing Interests:
I have read this submission. I believe that I have an appropriate level of expertise to confirm that it is of an acceptable scientific standard, however I have significant reservations, as outlined above. I want to thank all of you for the time and efforts spent to review our manuscript. All the authors of this paper are very grateful to have received such insightful comments from very observant scientists in the field. This certainly will help us to improve our manuscript. We have already started working on the second version of the paper. I will upload it on f1000-portal before the end of October and hope to address the concerns of all three reviewers.

Kind regards Vaibhav Bhatia
No competing interests were disclosed. Competing Interests: