A novel role for the DNA repair gene Rad51 in Netrin-1 signalling

Mutations in RAD51 have recently been linked to human Congenital Mirror Movements (CMM), a developmental disorder of the motor system. The only gene previously linked to CMM encodes the Netrin-1 receptor DCC, which is important for formation of corticospinal and callosal axon tracts. Thus, we hypothesised that Rad51 has a novel role in Netrin-1-mediated axon development. In mouse primary motor cortex neurons, Rad51 protein was redistributed distally down the axon in response to Netrin-1, further suggesting a functional link between the two. We next manipulated Rad51 expression, and assessed Netrin-1 responsiveness. Rad51 siRNA knockdown exaggerated Netrin-1-mediated neurite branching and filopodia formation. RAD51 overexpression inhibited these responses, whereas overexpression of the CMM-linked R250Q mutation, a predicted loss-of-function, had no effect. Thus, Rad51 appears to negatively regulate Netrin-1 signalling. Finally, we examined whether Rad51 might operate by modulating the expression of the Unc5 family, known negative regulators of Netrin-1-responsiveness. Unc5b and Unc5c transcripts were downregulated in response to Rad51 knockdown, and upregulated with RAD51 overexpression, but not R250Q. Thus, Rad51 negatively regulates Netrin-1 signalling, at least in part, by modulating the expression of Unc5s. Imbalance of positive and negative influences is likely to lead to aberrant motor system development resulting in CMMs.

The DNA repair gene RAD51 has been previously reported to maintain genetic stability during cell division through homologous recombination, by performing repair of double strand DNA breaks, and regulating cell cycle checkpoint mechanisms to prevent the replication of defective DNA and accumulation of mutations 1,2 . In vertebrate cells, Rad51 is essential for proliferation; cells deficient in Rad51 amass chromosomal abnormalities before eventual cell death 1 , and genetic deletion is embryonic lethal in mice 3,4 . Considering these known functions for maintaining DNA integrity, it was unanticipated when recent studies implicated mutations in RAD51 with both sporadic and familial cases of congenital mirror movements (CMM) [5][6][7] .
CMM are involuntary mirror reversals of movements occurring on one side of the body when voluntary movements on the opposite side of the body are performed, typically occurring in the hands 8,9 . Anomalous projections of motor systems are implicated in this disorder, for example, studies using focal transcranial magnetic stimulation (TMS) have revealed that in people with CMM, unilateral motor cortex stimulation evokes bilateral motor potentials in the muscles of the hands, suggesting abnormal ipsilateral corticospinal tract (CST) projections from motor cortex hand areas to the spinal cord (for review see 10 ). In addition, recording of scalp potentials during movement preparation, and interference with the cortical motor output by focal TMS, have shown bilateral activation of the motor cortex during intended movements, indicating abnormal callosal inter-hemispheric inhibition 10,11 . These findings indicate that CMM are likely underpinned by abnormal development of the decussating CST and/or corpus callosum [11][12][13] . In support of this, genetic analysis of people with CMM have identified mutations in Deleted in Colorectal Cancer (DCC) 5,14,15 , a receptor for Netrin-1, which is an axon guidance molecule known to be important for midline guidance decisions during development, including decussation of the CST and CC 16,17 as well as neuronal branching and filopodia formation of the axons that form these tracts 18 .
Dcc knockout mice display callosal agenesis 19 , and conditional Dcc mutants display a hopping gait, where the hind limbs are unable to move independently 20 , a motor phenotype that has been likened to mirror movements. Thus, whereas a deficit in DCC function correlates well with the features of CMM, the association with RAD51 is intriguingly less clear. The role of Rad51 in DNA repair in S and G 2 phases of division is well documented [21][22][23][24] , however there is no described function in post-mitotic, G 0 phase, terminally differentiated neurons, nor a role in motor system development. Although Rad51 can be detected in the brain throughout development, and is expressed within neural progenitor cells in the developing mouse neocortex from as early as E12 through to post natal stages 6 , it is unknown whether Rad51 serves any additional cellular function outside of DNA repair.
In light of the association of Rad51 mutation with CMM, we hypothesised that Rad51 may play a novel, as yet undescribed role in Netrin-1 signalling in the development of post-mitotic neurons within the motor system.

Results
Rad51 was redistributed within neuronal cytoplasm in response to Netrin-1. Genetic screening of people with CMM has identified mutations in RAD51 6,7 , or missense/truncating mutations in DCC, a Netrin-1 receptor 14,15 . The shared CMM phenotype from mutations affecting either of these genes lead us to hypothesize that RAD51 may be involved in Netrin-1 signalling. To investigate this, Rad51 subcellular distribution and expression levels were examined in primary neuron cultures from E14.5 mouse motor cortex using a Rad51 antibody, and co-labelling with neuron-specific beta-III tubulin (TuJ1). Rad51 immunoreactivity showed strong perinuclear aggregation in all TuJ1-positive, post-mitotic neurons (Fig. 1A,B). This contrasts with previous reports that describe a diffuse distribution of Rad51 through the cytoplasm and/or nucleus of neuronal, post-mitotic cells 6,25 .
Next, we examined neuronal Rad51 localisation following the application of recombinant mouse Netrin-1 (200 ng/mL) or vehicle (PBS). Netrin-1 treatment stimulated a redistribution of Rad51 protein within the neuronal cytoplasm, away from the border of the nucleus and extending distally down the primary axon at both 2 hr and 24 hr post-Netrin-1 treatment (Fig. 1C-F). This change was absent in PBS-treated neurons, which retained a perinuclear pattern of Rad51 expression (Fig. 1C,D). Significant redistribution of Rad51 within the neuronal cytoplasm could be observed after 2 hr Netrin-1 treatment, with Rad51 immunolabel detected nearly twice the distance from the nucleus compared to PBS-treated neurons (Fig. 1C  Twenty-four hours after Netrin-1 treatment, significant Rad51 redistribution away from the nucleus persisted ( Fig. 1D,F,G; Netrin-1: 15.11 μ m ± 0.860, n = 166; PBS: 8.414 μ m ± 0.473, n = 119, p < 0.0001; Unpaired t test with Welch's correction), extending further down the axon compared with 2 hr post-Netrin-1 (Fig. 1E,F,G, p < 0.05).
Quantification with qPCR showed no difference in Rad51 mRNA expression between Netrin-1-and PBS-treated cortical cultures at 2 hr or 24 hr, and no change to Rad51 protein abundance was detected by Western blotting after 24 hr Netrin-1 treatment (see Supplementary Fig. S1), indicating that a redistribution of existing cytoplasmic stores of Rad51 occurs in response to Netrin-1, as opposed to stimulation of transcription or translation. Together, these observations suggest a novel, non-canonical function for Rad51 in post-mitotic cortical neurons associated with Netrin-1 signalling, and distinct from its role in facilitating DNA damage repair in mitotic cells.
Knockdown of Rad51 in primary cortical neurons exaggerated Netrin-1-induced branching. Previous studies have shown that application of Netrin-1 to cortical neuron cultures results in increased filopodia formation and axonal branching 26,27 . To determine whether Rad51 might be involved in these Netrin-1-induced changes we performed siRNA-mediated knockdown of Rad51, and examined axon length, neurite number, and filopodia formation in both the presence and absence of Netrin-1 for 24 hr.
To confirm knockdown of Rad51, we first screened the efficacy of several siRNAs targeted to either the 5′ UTR or coding region of Rad51, and validated their ability to reduce the levels of Rad51 mRNA and protein in cultured cortical neurons (see Supplementary Fig. S2). This analysis also allowed us to further validate the specificity of the Rad51 antibody. After selection of the most effective Rad51 siRNA, E14.5 cortical neurons were treated with either siRNA specific for Rad51 (Rad51 siRNA), a scrambled control siRNA (scram siRNA), or Lipofectamine transfection reagent alone (vehicle). After 2 DIV, which was sufficient time for significant siRNA-mediated reduction of Rad51 mRNA and protein (see Supplementary Fig. S2), cortical cultures were exposed to either Netrin-1 or PBS for 24 hr, and axon length, branching, and filopodial formation were assayed.
Consistent with other in vitro studies 12,18,26 , axon length was not altered in response to Netrin-1 in any treatment group, nor was there a difference in total primary axon length between Rad51siRNA-treated neurons and scramsiRNA or vehicle controls ( Fig. 2A-D and I) at baseline (i.e., without Netrin-1) or in the presence of Netrin-1. Cell-staging analysis verified that the ability of cells to form an axon did not differ across treatment groups (see Supplementary Fig. S3). Thus, depletion of Rad51 mRNA and protein does not interfere with axonal outgrowth.
In the scram siRNA and vehicle control groups, 24 hr Netrin-1 treatment increased the number of neurites in comparison to PBS-treated control cells ( Fig Similarly, filopodia formation in response to Netrin-1 was also significantly increased in cultures treated with Rad51 siRNA compared to scram siRNA (p < 0.001), and vehicle (p < 0.001) (Fig. 2F,H,K). Moreover, in the absence of Netrin-1, Rad51siRNA-treated neurons exhibited a greater number of neurites (p < 0.05) and filopodia (vehicle < 0.001) than either scram siRNA or vehicle alone. These data suggest that Rad51 may be a negative regulator of Netrin-1 signalling in the modulation of axonal branching and filopodia formation.
Overexpression or knockdown of Rad51 in cortical neurons altered the expression of Netrin-1 receptors. Given that manipulation of Rad51 expression impacted the branching of neurons in response to Netrin-1 in a manner that suggests it plays a role in opposing the stimulatory effects of Netrin-1 on neuritogenesis, we examined whether this might be due to alterations in the expression of Netrin-1 receptors. Whereas Dcc typically has been reported to mediate the stimulatory effects of Netrin-1 on neuritogenesis. the Unc5 family We also quantified relative mRNA levels of Unc5a-d and Dcc in neurons transfected with EYFP alone, wtRAD51, or R250Q (Fig. 4B). In comparison to neurons expressing EYFP alone, overexpression of wtRAD51 increased the expression of Unc5b and Unc5c mRNAs ( Fig. 4B; Unc5b;1.57 ± 0.123 fold change, n = 8, p < 0.0001, Unc5c; 1.67 ± 0.24 fold change, n = 7, p < 0.05). There was no change in Unc5a (0.88 ± 0.114 fold change, n = 8, p > 0.05), Unc5d (1.19 ± 0.163 fold change, n = 10, p > 0.05), or Dcc mRNA (1.08 ± 0.040 fold change, n = 6, p > 0.05) in comparison to control. Interestingly, the relative mRNA expression of Unc5a-d and Dcc in neurons transfected with the R250Q construct did not differ from EYP control (Unc5a 1.031 ± 0.05990 n = 5, Unc5b 1.070 ± 0.1386 n = 5 Unc5c 1.120 ± 0.1113 n = 5 Unc5d 0.9328 ± 0.07761 n = 5, p > 0.05), or Dcc (0.9413 ± 0.02671 n = 5, p > 0.05), indicating that any regulatory function of Rad51 on Unc5 receptor expression may be lost as a result of the R250Q mutation.

Discussion
The DNA repair gene RAD51 acts to maintain genetic stability and prevent the replication of defective DNA in dividing cells. Curiously, mutations in RAD51 have recently been linked to CMM 5-7 , a developmental disorder of the motor system previously only attributed to deficits in Netrin-1/DCC signalling. Based on this we hypothesized that Rad51 is functionally involved in Netrin-1's effects on motor system development. We tested this hypothesis using a primary culture model of embryonic mouse neurons from motor cortex, which we challenged with Netrin-1 under conditions in which Rad51 expression and functionality were modulated. Our results indicate that Rad51 serves a novel, non-canonical function in neurons of the motor system, and that this role is in part linked to Netrin-1 regulation of axon branching. Although Rad51 expression can be detected in neuronal and non-neuronal cells in the brain, its cellular distribution has been described as weak or diffuse throughout the nucleus and/or cytoplasm 6,30 . In contrast, we observed Rad51 residing cytoplasmically in a perinuclear location within cortical neurons in vitro. In the developing brain, Rad51 is very strongly expressed in the nuclei of proliferative cells 6 , thus detection of lower levels of cytoplasmic Rad51 may have been difficult in tissue. In addition, we observed that Netrin-1 triggered Rad51 to redistribute within the cytoplasm from a perinuclear aggregation to a more dispersed distribution extending distally down the axon, with no concomitant changes in neuronal levels of Rad51 protein or mRNA. These observations suggest that in cortical neurons, a reorganization of existing stores of Rad51 protein occurs in response to Netrin-1. This suggests a functional link between Rad51 and Netrin-1, and further supports the hypothesis.
In light of the reported CST and/or CC tract malformations in CMM patients 11 , and Rad51 responsiveness to Netrin-1, we examined further whether Rad51 is involved in axonal branching, filopodia growth, or axon morphology of neurons from motor cortex, and whether such activities might be functionally linked with Netrin-1-mediated axon development. Interestingly, siRNA knock-down of Rad51 in the absence of Netrin-1 resulted in increased branching and filopodia formation to approximately the same extent as Netrin-1 in controls. When Rad51 was knocked-down in the presence of Netrin-1 however, a synergistic response was observed, where neurons with reduced expression of Rad51 displayed exaggerated branching in response to Netrin-1. Conversely, overexpression of RAD51 was found to have the opposite effect to knockdown, with a decrease in Netrin-1-induced branching. Neurons overexpressing Rad51 did not differ in the baseline number of neurites or filopodia when compared to controls, however in the absence of Netrin-1, neurons exhibited relatively few neurites and filopodia, so any further decrease may be biologically negligible and/or technically difficult to detect. These data suggest that Rad51 may act as a negative regulator of Netrin-1 signalling to modulate axonal branching and  filopodia formation of cortical neurons in the motor cortex. Additionally, RAD51 containing the R250Q mutation failed to inhibit Netrin-1 induced branching, supporting bioinformatics predictions that the mutation results in a loss of function. A loss of negative regulation of Netrin-1 signalling would be predicted to cause increased axonal branching of neurons, following exposure to Netrin-1 in vivo, leading to anatomical aberrations in the motor system. Functionally, this could result in reduced interhemispheric inhibition between cortical motor areas, lead to the persistence of both contralateral-and ipsilateral-projecting CST branches, or otherwise perturb the balance between side-specific activation and inhibition of a motor programme, resulting in the unwanted movements characteristic of CMM. Neurons transfected with the R250Q construct did not show an increase in neurite or filopodia formation above EYFP-controls, which suggests that the mutation specifically affects Rad51's Netrin-1-responsiveness, rather than a generic function in axon dynamics. Furthermore, the R250Q mutation is associated with CMM patients who present with no comorbidities, so the data suggests that the loss of Rad51 function associated with the R250Q mutation pertains specifically to its role in Netrin-1 signaling, and not its DNA repair functions.
In an attempt to ascertain a mechanism by which Rad51 might be having its effects on Netrin-1 signaling, we examined whether it might regulate the expression of genes encoding known stimulatory and inhibitory mediators of the Netrin-1 effects on axon dynamics. To this end, we have shown that neurons under-or over-expressing Rad51 exhibit changes in their mRNA expression profile of the Unc5 family of receptors, indicating a potential mechanism for Rad51 modulation of Netrin-1-induced branching. Specifically, Rad51 knockdown downregulated Unc5b and Unc5c receptor mRNA, whereas overexpression of wildtype RAD51 upregulated these receptor mRNAs. The increase in Unc5b and Unc5c transcript was absent in neurons overexpressing the R250Q construct, consistent with this mutant lacking Netrin-1 responsiveness. Owing to cellular heterogeneity (only a subpopulation of cortical neurons at this age would be Netrin-1-responsive), and transfection efficiency less than 100%, we believe that the somewhat modest expression changes we observe in vitro are likely to be an underestimate of what happens in a single Netrin-1-responsive cell in vivo.
Members of the Unc5 family of receptors are repulsive receptors for Netrin-1, negatively regulating filopodia dynamics and mediating growth cone collapse, with knockdown of either Unc5b or Unc5c associated with axon misguidance phenotypes [31][32][33] . In the developing cortical plate, Unc5c is expressed in a high-lateral to low-medial gradient to facilitate guidance of the interhemispheric projection 34 . In addition, Netrin-1 signalling via the Un5c receptors is important for the correct targeting of the CST, at the pyramidal decussation where CST axons traverse contralaterally in the caudal medulla, and at the dorsal funiculus in the spinal cord 20 . Thus, if Rad51 haploinsufficiency in CMM is similarly associated with disruption to Unc5 expression, then incorrect guidance of the CC and/or CST axons across the midline during development is likely to result. While there is no information available on Unc5 receptor mRNA levels in people with CMM affected by a Rad51 mutation, this would be a valuable area for further study.
Initially it may seem counterintuitive that mutations in either DCC or RAD51 can result in a CMM phenotype, given that they appear to work in an opposing manner on Netrin-1 signaling 35 . Formation of the CC is reliant on both attractive and repulsive signals however, and recent work suggests that Unc5c and Dcc might be required at different developmental timepoints for the correct guidance of callosal axons. For example, early pioneer axons from the cingulate cortex express Dcc and are attracted by Netrin-1 at the midline to cross into the contralateral hemisphere, whereas later born neurons from Layer V of the cortex express Unc5c and are repelled away from Netrin-1 at the internal capsule 34 , thus both attractive and repulsive signalling contribute to the midline crossing of callosal axons.
In conclusion, we have identified a novel role for Rad51, previously known only for its role in DNA repair. Our data suggest that Rad51 acts as a negative regulator of Netrin-1 signalling during development; a function that is consistent with observations in humans that RAD51 haploinsufficiency can lead to developmental defects in control of bilateral movements. These results add another player to the collection of positive and negative factors that work in concert to orchestrate proper neural development.

Methods
Cell Culture. All procedures described in this study were approved by the University of Otago, Animal Ethics Committee in accordance with the guidelines on Care and Use of Laboratory Animals (NIH Publication No. 85-23, 1996), and in accordance with the ARRIVE guidelines. All experiments were performed on C57BL/6J mice.
Timed-pregnant dams were sacrificed by cervical dislocation on E14.5 (with day of plug defined as E0.5), embryos were removed and decapitated in HBSS on ice. Sections of anterior cortices corresponding to the primary motor area were dissected and pooled within litters. Cells were washed with HBSS and dissociated with 50 μ L 0.5% Trypsin, 0.25% EDTA at 37 °C for 15 minutes. Dissociation was ceased by addition of 100 μ L FBS and cells were pelleted and resuspended in Neurobasal Media (no L-glutamine, no Phenol red) supplemented with 2% B27, 0.5 mM Glutamax and 0.1% Penicillin/Streptomycin. Cells were cultured in 24 well plates containing 13 mm round coverslips coated with sterile filtered 50 μ g/mL Poly-D-Lysine at a density of 10 −5 per well, and maintained at 37 °C, 5% CO 2 , 60% humidity. For Netrin-1 assays, recombinant mouse Netrin-1 (R&D systems), or PBS control, was diluted in plating media and added drop-wise to wells at a final concentration of 200 ng/mL. Reconstituted Netrin-1 was stored at − 80 °C and used within 3 months of reconstitution.
RNAi. All siRNA transfections were performed using RNAiMAX (Invitrogen) according to the manufacturer's instructions. Knockdown of Rad51 was performed using a combination of two siRNAs (10 nm each per well); coding region Rad51; 5′ -GGUAAAUCACCAACCAGGUAGU-3′ , 5′ UTR Rad51; 5′ -GUGGC CAGGAAGACUGAAAUU-3′ , Scramble control; 5′ -GGCAGACUACGUUACUACAGA-3′ . We initially screened three different siRNAs both individually and in combination, targeted to either the 5′ UTR or coding region of Scientific RepoRts | 7:39823 | DOI: 10.1038/srep39823 Rad51, validated their ability to reduce the levels of Rad51 mRNA and protein, and based on this, selected a duplex which gave the most significant reduction of Rad51 expression (see Supplementary Fig. S2). At 4 hours after plating, E14.5 cortical neurons were treated with either siRNA specific for Rad51 (Rad51siRNA), a scrambled control siRNA (scramsiRNA), or transfection reagent alone. Netrin-1 assays were performed 2 days in vitro (DIV) post-siRNA, which was determined to be sufficient time for significant siRNA-mediated reduction of Rad51 mRNA and protein (see Supplementary Fig. S2).

Rad51 Overexpression Assays.
The RAD51 open reading frame was PCR-amplified from human cDNA and cloned into a plasmid vector allowing the expression of an EYFP-RAD51 fusion transcript in mammalian cells (see Supplementary data for details). Sequence integrity and orientation was confirmed before large scale plasmid DNA amplification. Plasmid DNA was extracted by alkaline lysis, and DNA was eluted in 10 mM TE pH 8.0. Neurons were transfected at 1 DIV with either EYFP-RAD51, or EYFP-R250Q DNA constructs, or EYFP controls (500 ng/well), or vehicle alone, using Lipofectamine 2000 diluted in plating media. After 24 hours cultures were assayed for EYFP fluorescence.

RNA Extraction and Quantitative PCR (qPCR). Total RNA was extracted using Trizol (Gibco) and
Direct-zol RNA MiniPrep kit (Zymo), followed by Dnase-treatment using Ambion DNA-free DNase. RNA purity and quality was analysed using a NanoDrop ® ND-1000 spectrophotometer, and quantitated on a Qubit ® 2.0 Fluorometer using Qubit ® RNA Assay Kit. RNA was reverse transcribed in 25 μ l reactions using Superscript III reverse transcriptase as per manufacturers instructions. Resulting cDNA was stored at -80 °C until required. All primers for qPCR analysis were designed using PrimerBLAST (http://www.ncbi.nlm.nih.gov/tools/primerblast), sequences are listed in supplementary methods. All qPCR reactions were performed in triplicate in 10 μ L volume containing 2X SYBR Master mix (Bioline) and 1 μ L of cDNA template under standard thermal cycling conditions on a Viia7 Real Time PCR System (Applied Biosystems). No template control and no RT control reactions were included in each qPCR run, and gene expression was normalised to two reference genes: TATA box binding protein (Tbp) and phosphoglycerate kinase 1 (Pgk1). Fold change was determined in comparison to EYFP transfected controls using the 2− Δ Δ CT relative quantification method 36 . Immunocytochemistry. Immunocytochemistry was performed using standard protocols. Briefly, cells were fixed in 4% Paraformaldehyde in PBS, permeabilised with 0.25% Triton-X-PBS, blocked in 2% serum, 2% BSA, 0.1% Triton-X-PBS for 30 min, then incubated in primary antibody overnight at 4 °C. For detection of primary antibodies, cells were incubated with species-specific Alexa Fluor 488-or 594-conjugated secondary antibodies (1:500) for 90 min, and nuclei were counterstained with DAPI. Antibodies used were: rabbit anti-Rad51, 1:1000 (Abcam, ab63801), rat anti-β -III tubulin (TuJ1, monoclonal; R&D Systems) chicken anti-TuJ1 (Abcam, ab107216). Data analysis/Statistics/Imaging/NeuronJ. Neuron tracings were performed using the NeuronJ plugin 37 for ImageJ software (v1.46r, National Institutes of Health). Neurites were deemed to be the primary axon if they were at least twice the length of the cell body diameter, and the longest process. For secondary neurite measurements, data were separated into two categories: ≥ 5 μ m = neurites, < 5 μ m = filopodia. Comparisons between groups was performed using One-way ANOVA followed by Tukey's multiple comparison test. Confocal images were acquired on a Zeiss LSM710 microscope with a 63X objective, acquiring optical sections of 1 μ m.