RAD52 is required for RNA-templated recombination repair in post-mitotic neurons

It has been long assumed that post-mitotic neurons only utilize the error-prone non-homologous end–joining pathway to repair double-strand breaks (DSBs) associated with oxidative damage to DNA, given the inability of non-replicating neuronal DNA to utilize a sister chromatid template in the less error-prone homologous recombination (HR) repair pathway. However, we and others have found recently that active transcription triggers a replication-independent recombinational repair mechanism in G0/G1 phase of the cell cycle. Here we observed that the HR repair protein RAD52 is recruited to sites of DNA DSBs in terminally differentiated, post-mitotic neurons. This recruitment is dependent on the presence of a nascent mRNA generated during active transcription, providing evidence that an RNA-templated HR repair mechanism exists in non-dividing, terminally differentiated neurons. This recruitment of RAD52 in neurons is decreased by transcription inhibition. Importantly, we found that high concentrations of amyloid β, a toxic protein associated with Alzheimer's disease, inhibits the expression and DNA damage response of RAD52, potentially leading to a defect in the error-free, RNA-templated HR repair mechanism. This study shows a novel RNA-dependent repair mechanism of DSBs in post-mitotic neurons and demonstrates that defects in this pathway may contribute to neuronal genomic instability and consequent neurodegenerative phenotypes such as those seen in Alzheimer's disease.

The most deleterious form of DNA damage, double-strand breaks (DSBs), 2 can arise from endogenous metabolic pro-cesses or exogenous environmental factors such as radiation or chemicals (1). Oxidative damage caused by reactive oxygen species (ROS) and consequent DSBs has been implicated in neurodegenerative disorders such as Alzheimer's disease (AD). ROS are generated endogenously by cellular metabolism and a variety of exogenous agents, such as ionizing radiation (IR). In studies conducted on replicating cells, it has been found that metabolically generated ROS can cause around 10,000 oxidative lesions per day (2).
ROS-induced damage predominantly leads to base or deoxyribose modifications that lead to single-strand breaks (SSBs). DSBs can arise because of replication past ROS-induced lesions or when SSBs occur in close proximity (3). These lesions are repaired either via non-homologous end joining (NHEJ) or homologous recombination (HR). In NHEJ, the broken DNA ends are processed/digested and then directly ligated, potentially leading to nucleotide deletions and consequent frameshift mutations. Thus, NHEJ is considered to be error-prone (4). HR utilizes undamaged DNA templates to direct error-free repair of the damaged strands. It has long been assumed that HR can only take place during the late S/G 2 phases of the cell cycle, where sister chromatids are present as templates (5). However, recent studies in terminally differentiated cells have revealed an RNA-templated HR repair mechanism of DSBs at active transcription sites during the G 0 /G 1 phase (transcription-coupled homologous recombination (TC-HR)). For this mechanism to occur, Cockayne syndrome B protein is recruited to an actively transcribed damage site, followed by recruitment of replication protein A (RPA), RAD51, RAD51C, and RAD52; then, repair occurs utilizing the nascent RNA template produced by active transcription (6). An RNA-templated, RAD52-directed mechanism of HR has also been reported in yeast (7). In addition, a recent study identifies yeast and human RAD52 inverse strand exchange, in which RAD52 forms a complex with dsDNA and promotes strand exchange with homologous ssRNA or ssDNA, This work was supported in part by National Institutes of Health Grants GM118833 (to L. L.) and DC013048 (to M. E. R.). The authors declare that they have no conflicts of interest with the contents of this article. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health. This article contains Figs. S1 and S2. 1 To whom correspondence should be addressed: Dept cro ARTICLE indicating that RAD52 plays an important role in RNA-templated DSB repair in vivo (8). However, it is not known whether the transcription-dependent HR pathway exists in neurons. Except in the hippocampus and striatum, most neurons in the adult human brain are terminally differentiated and non-dividing and therefore are assumed to be incapable of utilizing HR for DSB repair (9,10).
In this study, utilizing our site-specific damage induction systems, we measured the recruitment of the TC-HR-associated repair protein RAD52 to sites of DNA damage with and without transcription inhibition. We found that post-mitotic neurons employ this RNA-based recombinational mechanism for the DNA damage response. We also discovered preferential binding of RAD52 protein to R-loops, DNA-RNA hybrid structures present during active transcription (11), as further evidence of the role of RAD52 in TC-HR. Given our finding of the novel TC-HR DNA repair mechanism in postmitotic cells, we wondered whether this mechanism might be affected in neurodegenerative disorders such as AD. We utilized A␤ 1-42 , neurotoxic oligomers 42 amino acids in length that are heavily implicated in AD pathology (12). We found that A␤ 1-42 oligomers down-regulate the expression and damage response of the essential TC-HR repair protein RAD52. How this dysregulation of DNA repair may significantly contribute to the development of neurodegenerative diseases such as AD is discussed.

DNA repair factors are recruited to laser-induced damage in neurons
To understand how non-dividing neurons repair oxidative DNA damage, we first verified that the primary rat cortical neurons utilized for experimentation were post-mitotic. Some findings have shown a link between cell cycle activation in postmitotic neurons and a DNA damage response leading to apoptosis (13). Therefore, we probed day in vitro 12 (DIV12) neuronal cell cultures with anti-phosphorylated histone H3 (Ser-10) before and after DSB-inducing ␥ irradiation (Fig. 1A). Studies have shown a correlation between this phosphorylated histone H3 (Ser-10) and mitotic chromosome condensation during early prophase, suggesting that anti-phosphohistone H3 can be used as a mitosis-specific marker (14). Here we observed no evidence of mitosis in primary cortical neurons compared with replicating U2OS cells before or 24 h after 5-gray ␥ irradiation.
Prior studies in our laboratory have shown that terminally differentiated cells utilize an RNA-templated homologous recombination mechanism for repair of DSBs in G 0 /G 1 phase of the cell cycle. This mechanism of repair can only occur with active transcription (6). We utilized 405-nm laser microirradiation to induce DSBs at localized sites in single cells, which makes it possible to observe repair protein recruitment in non-dividing cells (15) (Fig. 1B). We then investigated recruitment of GFP-tagged repair proteins specific to particular repair pathways, including XRCC1 (DNA single-strand break repair (SSBR)), KU70 (NHEJ), and RAD52 (HR) (Fig. 1C). In this experiment, we were surprised to find recruitment of the TC-HR repair protein RAD52 to sites of damage in post-mitotic neurons despite the lack of a sister chromatid of the neuron to utilize as a template for homologous repair.

DNA incorporation at sites of damage is affected by transcription inhibition
We next investigated whether transcription inhibition would affect the repair synthesis process of HR in post-mitotic neurons. We pretreated the cells with the RNA polymerase II inhib-itor5,6-dichloro-1-␤-D-ribofuranosylbenzimidazole(DRB),which inhibits transcription elongation, and analyzed the incorporation of bromodeoxyuridine (BrdU), a synthetic analog of thymidine, into newly repaired DNA after laser damage. Importantly, the replication-mediated incorporation of BrdU into genomic DNA of post-mitotic neurons does not occur. Therefore, BrdU staining in neurons should indicate the repair-triggered BrdU incorporation. This incorporation was significantly decreased after DRB treatment ( Fig. 1D), indicating that active transcription and a nascent RNA template are necessary for HR repair in post-mitotic neurons.

Transcription inhibition or RNase H treatment reduces the recruitment of RAD52 in neurons
RAD52 can be utilized in HR or a mutagenic repair process known as single-strand annealing. Single-strand annealing, which results in a deletion rearrangement between homologous repeating sequences, is independent of active transcription (16). Because our prior study indicated that the RAD52 dependent, RNA-templated recombination only occurs at sites of active transcription (6), we then tried to determine whether post-mitotic neurons utilize this particular recombinational mechanism for error-free repair of DSBs. To determine whether active transcription is necessary for the damage response of RAD52 in post-mitotic neurons, we pretreated cells with the RNA polymerase II inhibitor DRB, which inhibits transcription elongation, or ␣-amanitin, which inhibits transcription initiation and elongation (Fig. 2, A and B). Both inhibitors block transcription, thereby preventing the production of an mRNA template for use in RNA-templated HR. We found that, after inhibiting RNA polymerase II, recruitment of RAD52, an essential factor for RNA-templated recombinational repair, was significantly reduced at sites of laser damage in post-mitotic neurons. This supports our findings that RNA-templated HR in post-mitotic neurons requires active transcription to take place.
To further establish that the recombination in post-mitotic neurons is associated with RNA-templated repair, we examined the effect of RNase H, an endoribonuclease that specifically degrades the RNA strand in an RNA-DNA hybrid structure, on the recruitment of RAD52 (Fig. 2C). After treatment with RNase H, RAD52 recruitment was significantly inhibited at sites of laser damage in post-mitotic neurons. To test whether the effects of the RNase H assay were due to the enzyme activity or the treatment itself, we measured recruitment of the SSBR protein XRCC1 after RNase H treatment. Because XRCC1 does not rely on an RNA strand, recruitment of XRCC1 should not be affected The role of RAD52 in RNA-templated recombination by RNase H activity. As evidenced in Fig. 2D, we did not see a significant reduction in XRCC1 activity after RNase H treatment. Combined, these results confirm that a nascent RNA template is necessary for HR repair to occur in postmitotic neurons.

A␤ 1-42 sensitizes cells to IR in the absence of active transcription
The accumulation of A␤ in AD may be associated with DNA damage induced by oxidative stress. A recent study has shown that high concentrations of A␤  (1 M) in mitotic mouse hippocampal neuron cultures and in brain samples from AD patients contributed to reduced levels of BRCA1, leading to increased DSBs and reduced repair (17). However, it is not known whether A␤ 1-42 affects the major DSB repair pathways: the error-prone NHEJ pathway and the less error-prone HR pathway.
We investigated the effects of A␤ 1-42 on overall HR and NHEJ in dividing U2OS cells using the HR (DR-GFP) and NHEJ (Ej5-GFP) reporter assays (Fig. 3, A and C). For the HR reporter assay, the DR-GFP cell line utilizes an I-SceI recognition site inserted into a GFP coding sequence to create a nonfunctional GFP transgene. A wildtype GFP fragment has been inserted downstream of the transgene. After transfection of an I-SceIexpressing plasmid, a DSB is induced at the I-SceI cleavage site, and after HR occurs, fully functional GFP is restored, which can be detected by FACS. For the NHEJ assay, the Ej5-GFP cell line utilizes a GFP coding sequence interrupted by a puromycin gene that is flanked by I-SceI recognition sites. After transfection of the I-SceI-expressing plasmid, a DSB is induced at both sites, and, after NHEJ occurs, the fully functioning GFP is restored, which will be detectable by FACS (18,19). Both assays were treated with 1 M A␤ 1-42 to determine its effect on each pathway. A␤ 1-42 samples were investigated via Western blot- Error bars indicate the standard error of three separate experiments (n ϭ 100), and the p values were determined by using Student's two-tailed t test. Cortical neurons do not show expression of phosphorylated H3 (Ser-10). B, schematic of the 405-nm scan laser system for induction of DSBs in single cells, where single cell nuclei are targeted with 100 -500 ms of laser microirradiation to visualize and quantify GFP-tagged repair protein recruitment at damage sites. C, recruitment of the GFP-tagged DNA repair proteins XRCC1 (SSBR), KU70 (NHEJ), and RAD52 (HR) at sites of DNA damage in U2OS and DIV12 cortical neurons before and 1 min after 100-to 500-ms laser microirradiation. Error bars indicate the standard error of two separate experiments, and the p values were determined by using Student's two-tailed t test. Inc., increase. D, BrdU incorporation after 100-ms laser microirradiation with and without pretreatment with DRB (40 M) for 24 h in U2OS and DIV12 cortical neurons. Error bars indicate the standard error of three separate experiments (n ϭ 10), and the p values were determined by using Student's two-tailed t test (***, p Ͻ Ͻ 0.01). BrdU incorporation into repaired DNA was significantly reduced after transcription inhibition in post-mitotic neurons.

The role of RAD52 in RNA-templated recombination
ting and transmission electron microscopy to verify that they contained oligomers (Fig. S2, A and B). We found that A␤ 1-42 negatively affects classical HR in dividing cells (Fig. 3, B and D) to a slight extent.
To determine whether the slightly negative effect of A␤ 1-42 on classical HR was through the TC-HR pathway, we utilized the RNA polymerase II inhibitor DRB to disrupt transcription in U2OS cells and measured clonogenic survival against IR. 1 M A␤ 1-42 alone negatively affects overall U2OS survival (Fig.  3E), and, as shown in Fig. 3F, DRB treatment rendered cells more sensitive to IR. However, addition of 1 M A␤ 1-42 to the DRB treatment did not further increase the sensitivity of the cells to IR (Fig. 3G). This indicates that the effect of A␤ 1-42 on classical HR is dependent on active transcription.

High concentrations of A␤ 1-42 down-regulate the damage response of RAD52 in neurons
After determining that the RNA-templated recombination in neurons is active, we tested how A␤ 1-42 might compromise repair efficiency in non-dividing neurons. We treated primary rat cortical neuron cell cultures with 1 M A␤ 1-42 (Fig. 3, H and  I). Western blot analysis showed that 1 M A␤ 1-42 reduced overall RAD52 protein levels in post-mitotic neurons (Fig. 3H) and led to increased ␥H2AX expression (Fig. S1). After 24 h of treatment, it significantly reduced RAD52 recruitment to damage sites in both post-mitotic neurons and U2OS cancer cells (Fig. 3I). This indicates that A␤ 1-42 affects the same repair pathway in both dividing and non-dividing cells, compromising

The role of RAD52 in RNA-templated recombination
genomic stability in a similar manner by contributing to the reduction of TC-HR repair efficiency.

RAD52 binds to an R-loop substrate and preferentially binds to ssRNA
RAD52 protein has been demonstrated to bind DNA structures with DSBs (20,21). Because previous studies from others and our study both indicate that RAD52 is involved in RNA-templated recombination repair, we tested the affinity of RAD52 to RNA-related substrates by electrophoretic mobility shift assay. The results show that purified human RAD52 protein efficiently binds both ssDNA and RNA (Fig. 4, A and B), although its binding affinity to ssRNA is higher than that of ssDNA. The binding affinities of RAD52 to dsDNA and RNA (Fig. 4, C and D) are significantly reduced compared with single-stranded substrates (Fig. 4G). Interestingly, RAD52 has a

The role of RAD52 in RNA-templated recombination
slightly higher affinity to an RNA-DNA hybrid duplex than dsDNA and dsRNA (Fig. 4E). Furthermore, RAD52 has a higher binding affinity to R-loop substrates (Fig. 4F), the damageprone structures at transcriptionally active sites in the genome (11), than hybrid structures. These in vitro results are consistent with the possibility that RAD52 may bind ssRNA and the co-transcriptional R-loop upon transcriptional stress in postmitotic neurons.

Discussion
In this study, we first discovered that recruitment of the HR factor RAD52 to sites of DNA damage is dependent on the presence of a nascent mRNA template for HR, indicating the existence of a novel RNA-templated recombinational repair pathway in post-mitotic neurons. We also discovered that RAD52 preferentially binds ssRNA and has direct affinity for R-loops, reinforcing the involvement of RAD52 in TC-HR repair of double-strand breaks. Utilization of this less errorprone pathway is essential for non-dividing neurons to maintain their genomic integrity when faced with environmental pressures such as metabolically induced oxidative damage and other endogenous and exogenous stressors.
Considering the fact that terminally differentiated, post-mitotic neurons must survive for a human lifetime, utilizing an errorprone pathway such as NHEJ would be vastly insufficient. In a study on cell cycle-arrested cells, it was found that NHEJ activity was responsible for approximately half of all replication-independent frameshift mutations, primarily because of deletions. Mutants deficient for HR (RAD54 deletion) contributed no increase in the amount of replication-independent mutations in the cell cyclearrested cells but RAD52-deletion mutants did (22). This study confirms both the mutagenic nature of NHEJ and the possibility of a RAD52-dependent pathway that counteracts the frameshift-

The role of RAD52 in RNA-templated recombination
producing NHEJ pathway, such as TC-HR, which we have elucidated here and in our previous work.
Although post-mitotic neurons do not face the acquisition of mutations after replication past DNA lesions, DNA DSBs still pose massive risks to essential genes when not repaired. Failure to completely repair DNA may result in cellular senescence or apoptosis as a means for the cell to cope with an excess of damage and avoid malignant transformation (23,24). However, unrepaired cells may also accumulate mutations that affect genes involved in regulating apoptosis, cell division, and DNA repair. If proliferating stem and neural progenitor cells incur mutations because of unrepaired DNA during development, then disease may later develop as a result of disrupted essential genes passed down through cell division or when tumor suppressing or oncogenes are no longer properly regulated (25). In post-mitotic neurons, mutations may contribute to neuronal dysfunction and apoptosis, as seen in the pathology of AD, where the cerebral cortex and hippocampus suffer extensive neuronal loss (26). Many neurodegenerative pathologies have also been linked to genetic deficiencies in DNA repair genes. For example, in ataxia telangiectasia, a mutation in the PI3K ataxia telangiectasia-mutated (ATM) protein, which is activated by DNA DSBs and is responsible for phosphorylating numerous downstream targets involved in DNA repair, cell cycle arrest, and apoptosis, causes individuals to experience higher rates of cancer and undergo neurodegeneration before the age of five (27)(28)(29). New research has suggested that the neuronal loss observed in AD may also result from partial loss of ATM function as AD pathology progresses (30).
In this study, we also found that high concentrations (micromolar) of A␤ 1-42 oligomers lead to a reduction in the key RNAtemplated HR repair protein RAD52. Given the critical role of RAD52 in the RNA-templated recombination process but not classic recombination in S/G 2 phase, this reduction in overall function may result in increased DSBs because of a loss in highfidelity repair capacity. The question remains, however, of the direct cause of increased concentrations of A␤ 1-42 found in AD patients and how they directly or indirectly down-regulate essential repair proteins. Whether deficiencies in DNA repair, genetic abnormalities, or exogenous factors are the source of this pathology is still under investigation.
Study of repair mechanisms in post-mitotic neurons is essential if we are to understand the complex nature of neurodegenerative diseases. Although certain genetic aberrations may constitute a significant role in contributing to physiological aspects of AD, such as the APOE⑀4 allele and its association with increased AD risk, other contributing factors remain unknown (31). These factors may include deficiencies in DNA repair directly related to environmental exposure or internal processes such as the production of A␤ 1-42 .

A␤ 1-42 oligomer preparation
Recombinant A␤ 1-42 oligomers were prepared from lyophilized monomers treated with hexafluoro-2-propanol (HFIP) (A-1163-2, rPeptide) as described previously (17,34). A␤ was resuspended in DMSO at 1 mg/ml and further diluted with ice-cold Neurobasal medium to 100 -200 M. Protein concentration was measured using reverse-phase HPLC and the Pierce BCA protein assay kit (Thermo Scientific). The ratio of monomers versus oligomers was determined using size exclusion chromatography on a sample incubated for 24 h. We also investigated the pelleted and supernatant fractions and observed that the supernatant fraction was the toxic fraction and that DMSO did not induce the resultant effects. Additionally, we investigated A␤ 1-42 oligomers via Western blotting by running diluted fractions with monomeric controls on a 4 -20% gradient gel (Mini-Protean TGX, 4561096) and probing with anti-6E10 antibody (BioLegend, 803004) (Fig. S2A). A␤ control The role of RAD52 in RNA-templated recombination medium was generated using equivalent amounts of DMSO and Neurobasal medium.

Transmission electron microscopy
Transmission electronic microscopy was used to detect the presence of A␤ 1-42 oligomers (Fig. S2B). The A␤ 1-42 sample (5 l) was placed on a formvar grid and allowed to partly dry at room temperature. It was washed by touching it three times to the surface of a drop of distilled water, and excess water was removed by touching the grid to filter paper. Then a small drop of 1-3% uranyl acetate in distilled deionized water was added to the grid. After 10 s, excess stain was removed by touching the edge to filter paper. The grid was allowed to air-dry at room temperature before examination using a transmission electron microscope (Jeol JEM 1400, Tokyo, Japan) and imaging at 80 kV with the MegaView III Soft Imaging System.

Polymerase II inhibitors
The RNA polymerase II inhibitor DRB (D1916, Sigma) was added with a final concentration of 40 M for 24 h, or ␣-amanitin (A2263, Sigma) was used at 100 g/ml for 15 min before laser microirradiation.

RNase H treatment
U2OS cells and neurons were rinsed once with PBS and incubated with 1ϫ RNase H buffer with or without RNase H (EN0201, Thermo Scientific) at room temperature for 15 min. Cells were treated with 0.1 mM 8-MOP (Methoxsalen) and incubated for 10 min at 37°C, followed by confocal microscopy.

Microscopy and laser light irradiation
Fluorescence images were obtained and processed with an FV-500 confocal scanning laser microscopy system (Olympus). A 405-nm scan laser system (Olympus) for irradiation of cells in the epifluorescence path of the microscope system was used. One scan of the laser light at full power delivers ϳ1600 nanowatts. We scanned cells 100 -500 times with the 405-nm laser at full power focused through a ϫ40 objective lens, which has been shown to induce SSBs and DSBs. Live cells were treated with 0.1 mM 8-MOP in Opti-MEM (Gibco) in glass-bottom dishes, incubated for 10 min at 37°C, and then placed on a temperature-controlled (37°C) plate. "Before" images were taken prior to laser microirradiation. The fluorescence intensity at an irradiated site was initially measured with a laser power/energy monitor (Orion, Ophir Optronics). The -fold increase in fluorescence intensity of each site was quantified by measuring fluorescent intensity after damage induction/background nuclear intensity using ImageJ software (15,32,35). Each experiment was done at least twice, and we observed 10 or more cells to obtain one result. Standard deviations and standard errors were derived from at least two independent experiments, and statistical significance (p values) was assessed using an unpaired Student's t test: *, p Ͻ 0.05; **, p Ͻ 0.01; ***, p Ͻ Ͻ 0.01.

Protein expression and purification
His 6 -tagged human RAD52 was expressed in Escherichia coli Rosetta cells and purified as described previously (36).

HR and NHEJ assays
DR-GFP and Ej5-GFP stable cells were incubated in Dulbecco's modified Eagle's medium with 10% FBS. Cells were transfected with the pCMV-I-SceI plasmid using Lipofectamine 2000 (Life Technologies) for 24 h and kept in the dark. 1 M A␤ 1-42 or A␤ control medium was added for 5 h. Then cells were spun at 600 rpm, washed with 1ϫ PBS three times, and analyzed by FACS.

BrdU incorporation
U2OS or DIV12 rat cortical neurons previously transfected with GFP-XRCC1 were treated with 40 M DRB for 24 h or left untreated. Cells were treated with laser microirradiation, and then 0.01 mM BrdU was added for 24 h. Cells were fixed with 3.7% paraformaldehyde for 15 min at room temperature, rinsed The role of RAD52 in RNA-templated recombination three times with PBS, and then treated with 2.5 N HCl for 30 min at 37°C. Cells were rinsed three times with PBS, permeabilized with 0.2% Triton for 5 min at room temperature, and then rinsed three times with PBS. Immunofluorescence staining proceeded as above.

Cell survival
To determine the effect of 1 M A␤ 1-42 on U2OS cell survival, U2OS cells were treated with 1 M A␤ 1-42 or A␤ control medium for 24 h or left untreated, seeded onto 6-cm dishes, and incubated at 37°C for 10 days. Colonies were fixed and stained with 0.3% crystal violet, and then survival (number of colonies) was expressed as a percentage of non-treated colonies. To determine the effect of 1 M A␤ 1-42 on U2OS IR sensitivity, U2OS cells were treated with 40 M DRB for 24 h and/or 1 M A␤ 1-42 or A␤ control medium for 5 h or left untreated and then seeded onto 6-cm dishes. After seeding, cells were exposed to radiation as indicated and incubated for 10 days at 37°C. Colonies were fixed and stained with 0.3% crystal violet, and then survival (number of colonies) was expressed as a percentage of non-irradiated colonies.
Author contributions-S. W. conducted the major experiments and wrote the manuscript. Y. T., Z. L., W. Z., L. H. S., M. E. R., R. K., and A. T. helped with the assays. J. T. G., A. T., and R. W. helped analyze the results. A. S. L. and L. L. provided overall experimental guidance. All authors edited their specific contributions and reviewed the manuscript.