The RNF8 and RNF168 Ubiquitin Ligases Regulate Pro- and Anti-Resection Activities at Broken DNA Ends During Non-Homologous End Joining

The RING-type E3 ubiquitin ligases RNF8 and RNF168 recruit DNA damage response (DDR) factors to chromatin flanking DNA double strand breaks (DSBs) including 53BP1, which protects DNA ends from resection during DNA DSB repair by non-homologous end joining (NHEJ). Deficiency of RNF8 or RNF168 does not lead to demonstrable NHEJ defects, but like deficiency of 53BP1, the combined deficiency of XLF and RNF8 or RNF168 leads to diminished NHEJ in lymphocytes arrested in G0/G1 phase. The function of RNF8 in NHEJ depends on its E3 ubiquitin ligase activity. Loss of RNF8 or RNF168 in G0/G1-phase lymphocytes leads to the resection of broken DNA ends, demonstrating that RNF8 and RNF168 function to protect DNA ends from nucleases, pos sibly through the recruitment of 53BP1. However, the loss of 53BP1 leads to more severe resection than the loss of RNF8 or RNF168. Moreover, in 53BP1-deficient cells, the loss of RNF8 or RNF168 leads to diminished DNA end resection. We conclude that RNF8 and RNF168 regulate pathways that both prevent and promote DNA end resection in cells arrested in G0/G1 phase.


Introduction
DNA double strand breaks (DSBs) are repaired by either non-homologous end joining (NHEJ) or homologous recombination (HR) [1][2][3]. NHEJ functions in all phases of the cell cycle to rejoin broken DNA ends and is the main pathway of DSB repair in G 1 -phase cells and non-cycling G 0 cells [2]. In contrast, HR functions in the S-and G 2 -phases of the cell cycle using the sister chromatid as a template for precise DNA DSB repair [1,3]. DNA repair by either pathway relies on the recruitment of DNA damage response (DDR) factors to chromatin at DSBs [1,4]. Phosphorylation of the histone variant H2AX, forming γ-H2AX, in chromatin flanking DNA DSBs is an early step in the DDR that leads to the recruitment of several DDR factors including the RNF8 and RNF168 RING-type E3 ubiquitin ligases [5][6][7]. RNF168 ubiquitylates H2A and H2AX, promoting the recruitment of additional DDR factors, including 53BP1 and BRCA1 [5,6,[8][9][10][11][12][13][14][15]. 53BP1, along with downstream effectors RIF1 and the Shieldin complex, protects DNA ends from resection, which can lead to single stranded DNA (ssDNA) overhangs that antagonize repair by NHEJ [16][17][18]. In contrast, BRCA1 promotes DNA end resection, which is required for DNA DSB repair by HR [1,3]. How these activities are balanced to promote DSB repair by HR or NHEJ is not completely known.
The assembly of antigen receptor genes occurs in G 0 /G 1 -phase developing lymphocytes through V(D)J recombination, a reaction that is initiated when the RAG-1 and RAG-2 proteins, which together from the RAG endonuclease, introduce DNA DSBs at the border of two recombining gene segments [19]. DNA cleavage by RAG leads to the formation of a hairpin sealed coding end (CE) and a blunt signal end (SE) at each DSB. The two CEs are then joined to form a coding join (CJ) and the two SEs joined to form a signal join (SJ) via NHEJ [20]. NHEJ is dependent on a set of core factors, including KU70, KU80, DNA Ligase IV and XRCC4, with loss of any of these factors leading to a complete block in NHEJ and RAG DSB repair [2,20]. There are additional NHEJ proteins that are not absolutely required for DSB repair. One of these, the XRCC4-like factor (XLF), functions during NHEJ in non-lymphoid cells but is largely dispensable for NHEJ in lymphoid cells [21][22][23][24]. However, lymphocytes with combined deficiencies of XLF and other DDR proteins have shown demonstrable defects in the repair of RAG DSBs during V(D) J recombination and thus revealed the functions of XLF and these proteins in NHEJ-mediated DSB repair [25][26][27][28][29][30][31][32][33]. In this regard, while loss of either XLF or 53BP1 leads to only modest defects in RAG DSB repair in lymphocytes, the combined loss of XLF and 53BP1 leads to a dramatic block in the NHEJ-mediated repair of these DSBs [27].
Here we examine the function of RNF8 and RNF168 during NHEJ in wild type and XLFdeficient lymphocytes. These analyses reveal that both RNF8 and RNF168 function during NHEJ. Moreover, we show that, like 53BP1, RNF8 and RNF168 protect DNA ends from nucleolytic resection in G 0 /G 1 -phase cells. However, DNA end resection in 53BP1-deficient cells is more severe than in RNF8-or RNF168-deficient cells and cells with a combined deficiency of 53BP1 and either RNF8 or RNF168 exhibit diminished resection as compared to cells only deficient in 53BP1. Thus, RNF8 and RNF168 function to balance pro-and anti-resection activities at broken DNA ends in G 0 /G 1 -phase cells.

Southern blot analyses
10 μg of genomic DNA from abl pre-B cells with pMG-INV was digested with NheI or XbaI and resolved in 1.2% TAE gels. Upon transferring DNA to Zeta Probe (Bio Rad, 1620165) membranes, V(D)J recombination products and intermediates of pMG-INV were visualized using Thy1 probe (800 bp Thy1.1 cDNA) as previously described [26]. Quantification of the intact SEs in Lig4 −/− and other cell lines derived from Lig4 −/− was done using Image J and calculated as the percentage of the intensity of intact signal ends (iSEs) over the total intensity of unrearranged substrate (UR), iSEs and resected signal ends (rSE) (the region from 0.5 to 3.7 kb). Southern blot analyses were carried out on two independently isolated mutants of each genotype for at least twice.

Retroviral expression of WT and mutant RNF8
The murine RNF8 coding sequence was amplified from cDNA clone BC021778 (transOMIC) by PCR and cloned to pOZ-FH-N downstream of FLAG-HA epitope tag at XhoI and NotI sites. DNA sequences corresponding to amino acids 37-109 and 405-447 are deleted in mutant RNF8 (ΔFHA) and RNF8 (ΔRING), respectively.

Flow cytometric analysis of radiation-induced chromatin association of RPA
Analysis of DNA-damage induced RPA binding to chromatin was performed as previously described with modifications [39]. Briefly, abl pre-B cells were treated with 3 μM imatinib for 2 days, followed by 10 Gy irradiation. 8 hours after irradiation, cells were permeabilized in 150 μl of 0.05% Triton X-100/1X PBS on ice for 10 min and immediately diluted in 2 ml of FACS wash (2% fetal bovine serum (FBS)/1X PBS). Permeabilized cells were fixed in 150 μl of BD Cytofix/Cytoperm at room temperature for 15 min and then washed in 1 ml of FACS wash. Cells were then stained with anti-RPA antibody (Cell Signaling, 2208S) and Alexa Fluro 647 anti-rat IgG (Biolegend, 405416) and analyzed with an LSRII (BD Biosciences). Two independently isolated mutants of each genotype were used in the flow cytometric assays and repeated at least twice.

RNF8 functions during NHEJ
Murine Abelson virus-transformed pre-B cells (abl pre-B cells) were generated containing a single copy of the pMG-INV retroviral recombination substrate (Fig. 1A) [26]. pMG-INV has two recombination signal sequences (RSs) flanking an antisense GFP cDNA ( Fig. 1A) [26]. RAG cleavage at the pMG-INV RSs leads to the generation of a pair of signal ends (SEs) and coding ends (CEs) at each DSB (Fig. 1A). NHEJ then joins the two SEs to generate a signal join (SJ) and the two CEs to generate a coding join (CJ) leading to inversion of the GFP cDNA and GFP expression as an indicator of completed V(D)J recombination (Fig. 1A). The CE from one DSB can also be aberrantly joined to the SE from the other DSB to generate a hybrid join (HJ) (Fig. 1A). To induce V(D)J recombination, abl pre-B cells are treated with the abl kinase inhibitor, imatinib, which leads to G 0 /G 1 cell cycle arrest, induction of RAG and V (D)J recombination at pMG-INV [36] as indicated by robust GFP expression in WT cells (Fig. 1B).
RNF8-deficident abl pre-B cells were generated from Rnf8 −/− mice and pMG-INV was introduced into these cells to monitor V(D)J recombination and NHEJ [34] and treatment with imatinib leads to pMG-INV recombination equivalent to WT abl pre-B cells as indicated by GFP expression (Fig. 1B). Similar to what was observed in WT abl pre-B cells, Southern blot analyses revealed pMG-INV SJ formation with no significant accumulation of un-repaired SEs at RAG DSBs in Rnf8 −/− abl pre-B cells (Fig. 1C). We conclude that RNF8 deficiency does not lead to a demonstrable defect in RAG DSB repair during V(D)J recombination in abl pre-B cells.
We next generated abl pre-B cells from Xlf −/− and Xlf −/− :Rnf8 −/− mice, followed by integration of the pMG-INV reporter. Xlf −/− abl pre-B cells exhibit robust V(D)J recombination of pMG-INV, similar to WT abl pre-B cells as evidenced by GFP expression (Fig. 1B) and Southern blot analyses of SJ formation (Fig. 1C) [23]. In contrast, Xlf −/− :Rnf8 −/− abl pre-B cells exhibited a decrease in the fraction of GFP expressing cells after RAG induction, indicative of a defect in V(D)J recombination (Fig. 1B). Indeed, there was a reduction in the hybridization intensity of the SJ band in Xlf −/− :Rnf8 −/− abl pre-B cells as compared to the WT, Xlf −/− and Rnf8 −/− abl pre-B cells (Fig. 1C). Moreover, un-repaired SEs were detectable by Southern blot in Xlf −/− :Rnf8 −/− abl pre-B cells (Fig. 1C). Notably, the intensity of these SE bands diminishes over time, possibly reflecting the nucleolytic resection of the DNA ends (see below).
There are two conserved domains in RNF8, the FHA and RING domains [5,11]. The FHA domain promotes the association of RNF8 with phosphorylated MDC-1 at the site of a DSB and the RING domain is required for RNF8 E3 ligase catalytic activity [11,12].
To determine if these domains of RNF8 are required for NHEJ we expressed wild type RNF8 or mutants lacking the FHA (ΔFHA) or RING (ΔRING) domains in Xlf −/− :Rnf8 −/− abl pre-B cells (Fig. 1D). As compared to cells infected with empty vector, induction of V(D)J recombination in Xlf −/− :Rnf8 −/− abl pre-B cells expressing RNF8 resulted in a significant rescue of pMG-INV rearrangement as evidenced by GFP expression and Southern blot analyses ( Fig. 1E and F). However, expression of either the ΔFHA or ΔRING RNF8 mutants, even though their levels were elevated over that of WT RNF8 (Fig. 1D), failed to improve pMG-INV rearrangement. We conclude that RNF8 functions during NHEJ-mediated DSB repair in ways that depend on the FHA and RING domains ( Fig.  1E and F).

RNF8 and RNF168 are epistatic during NHEJ
RNF168 has functions downstream of RNF8 in response to DNA damage [6,8]. To determine if RNF168 also functions in NHEJ, Rnf168 −/− and Xlf −/− :Rnf168 −/− abl pre-B cells were generated from WT and Xlf −/− abl pre-B cells through the CRISPR/Cas9mediated inactivation of both Rnf168 alleles and the loss of RNF168 protein in these cells was confirmed by Western blot (Fig. 2A) (Fig. 3C). We conclude that RNF8 and RNF168 function in the same pathway during NHEJ-mediated DSB repair.

RNF8 and RNF168 inhibit DNA end resection in G 0 /G 1 -phase cells
RNF8 and RNF168 function to localize 53BP1 to DSBs and 53BP1 functions to protect DNA ends from resection [5,6,8,11,12,17]. To determine whether RNF8 and RNF168 function to protect DNA ends from resection in cells arrested in G 0 /G 1 phase, we used abl pre-B cell lines deficient in DNA Ligase IV (Lig4 −/− ) to generate cells deficient in DNA Ligase 4 and 53BP1 (Lig4 −/− :53bp1 −/− ), RNF8 (Lig4 −/− :Rnf8 −/− ) or RNF168 (Lig4 −/− :Rnf168 −/− ) by CRISPR/Cas9 inactivation of both 53bp1, Rnf8 or Rnf168 alleles, respectively (Fig. S1). Treatment of Lig4 −/− abl pre-B cells with imatinib led to RAG cleavage at pMG-INV, and RAG DSBs persist un-repaired due to the deficiency of DNA Ligase 4 and the lack of NHEJ in these cells (SE, Fig. 4A). The persistence of DSBs in these cells allows to better analyze DNA end processing in response to different genetic manipulations. The hybridizing band due to the un-repaired SEs in Lig4 −/− abl pre-B cells remained homogenous in size over four days indicating that these DNA ends were not significantly resected (Fig. 4A). In contrast, the hybridizing bands from SEs generated in Lig4 −/− :53bp1 −/− abl pre-B cells were significantly reduced in intensity, shown as the reduced percentage of intact SEs (iSEs) in these cells when compared with that in Lig4 −/− abl pre-B cells, and existed as a heterogeneously sized 'smear' of lower molecular weight bands indicative of extensive DNA end resection (rSEs) (Fig. 4A) [40,41].
The proportion of intact, unprocessed SEs in Lig4 −/− :Rnf8 −/− and Lig4 −/− :Rnf168 −/− abl pre-B cells is also lower than that in Lig4 −/− abl pre-B cells and heterogenous bands of small sizes could also be detected as Lig4 −/− :53bp1 −/− abl pre-B cells, suggesting the SEs in these cells also exhibited resection, although to a much lesser extent than what was observed in Lig4 −/− :53bp1 −/− abl pre-B cells (Fig. 4A). That these DNA ends were resected was further evidenced by the shRNA-mediated knockdown of CtIP, which is required for the resection of RAG DSBs in G 0 /G 1 -arrested abl pre-B cells ( Fig. 4B and C) [37]. Reduced expression of CtIP in Lig4 −/− :Rnf8 −/− and Lig4 −/− :Rnf168 −/− abl pre-B cells led to an increased intensity of the full length SE band and a reduction in the lower molecular weight smear, indicative of diminished resection of broken DNA ends ( Fig. 4D and E).
To assess whether the role of RNF8-and RNF168 in limiting DNA end resection extended beyond V(D)J recombination, we irradiated Lig4 −/− , Lig4 −/− :53bp1 −/− , Lig4 −/− :Rnf8 −/− and Lig4 −/− :Rnf168 −/− abl pre-B cells that had been arrested in G 0 /G 1 by treatment with imatinib and assayed for RPA association with chromatin by flow cytometry (Fig. 5) [39]. The RPA complex associates with single strand DNA (ssDNA) generated by nucleolytic resection at broken DNA ends and therefore the level of RPA associated with chromatin after DNA damage is indicative of ssDNA formation at DSBs [1]. After irradiation, Lig4 −/− abl pre-B cells exhibited much lower levels of chromatin-bound RPA when compared to Lig4 −/− :53bp1 −/− abl pre-B cells, consistent with the function of 53BP1 in protecting broken DNA ends from nuclease activity [16][17][18]41]. Irradiated G 0 /G 1 -arrested Lig4 −/− :Rnf8 −/− and Lig4 −/− :Rnf168 −/− abl pre-B cells also exhibited increased RPA association with chromatin indicative of the formation of ssDNA at irradiation-induced DSBs (Fig. 5). There was less chromatin-bound RPA in irradiated Lig4 −/− abl pre-B cells lacking RNF8 or RNF168 as compared to those lacking 53BP1, suggesting that there is more ssDNA at DSBs in cells lacking 53BP1 as compared to those lacking RNF8 or RNF168 (Figs. 4A and 5). We Chen et al. Page 6 conclude that RNF8 and RNF168 have functions that can inhibit DNA end resection at DSBs in G 0 /G 1 -phase cells at both RAG generated and irradiation induced DNA ends.

RNF8 and RNF168 promote DNA end resection in G 0 /G 1 -phase cells
RNF8 and RNF168 promote the association of 53BP1 with DSBs, yet we find less resection of DSBs (higher percentage of intact SEs) in G 0 /G 1 -phase cells deficient in RNF8 or RNF168 as compared to those deficient in 53BP1 (Figs. 4A and 5). This suggests that RNF8 and RNF168 may also have functions in promoting resection. To test this, we developed a CRISPR/Cas9 approach to efficiently inactivate both alleles of 53bp1, Rnf8 or Rnf168 in bulk populations of Lig4 −/− and Lig4 −/− :53bp1 −/− abl pre-B cells (Fig. 6A). A lentivirus encoding doxycycline inducible Cas9 was introduced into Lig4− /− or Lig4 −/− :53bp1 −/− abl pre-B cells followed by introduction of lentiviruses expressing guide RNAs (gRNAs) targeting 53bp1, Rnf8 or Rnf168. Treatment of these cells with doxycycline for 7 days led to significant loss of 53BP1 and RNF168 protein in the bulk populations, a process we termed bulk gene inactivation ( Fig. 6A and B).

Discussion
Here we show that the RNF8 and RNF168 RING E3 ubiquitin ligases function in NHEJmediated DSB repair regulating DNA end resection during V(D)J recombination and after irradiation in lymphocytes. Although RNF8 or RNF168 deficiency alone does not lead to a defect in V(D)J recombination, the combined deficiency of either of these proteins with XLF leads to a demonstrable defect in RAG DSB repair. Therefore, while RNF8 and RNF168 are not required for NHEJ under normal conditions, their roles in NHEJ can be revealed in the absence of XLF through a "synthetic NHEJ defect" resulting from combining the deficiency of non-essential functions of these proteins. The function of RNF8 in NHEJ requires both the FHA and RING domains suggesting that both the localization of RNF8 to DSBs through phospho-MDC1 (FHA domain) and its E3 ubiquitin ligase activity (RING domain) are important in this context. Moreover, our finding that these two proteins are epistatic with respect to their activities in NHEJ-mediated DSB repair is consistent with the requirement for RNF8 to recruit RNF168 to DSBs and raises the possibility that the primary role of RNF8 is in recruiting RNF168 [6,8].
How is it that RNF8 and RNF168 function during NHEJ? One explanation may be that they protect DNA ends until they can be joined by NHEJ. We find that the loss of RNF8 or RNF168 can lead to the resection of DNA ends generated by RAG cleavage in G 0 /G 1 -phase abl pre-B cells that are unable to repair DSBs due to deficiency of DNA Ligase 4. Moreover, irradiation of G 0 /G 1 -phase abl pre-B cells that are deficient in DNA Ligase 4 and either RNF8 or RNF168 leads to increased association of RPA with chromatin indicative of ssDNA formation at DSBs [2,42]. The formation of these ssDNA overhangs would antagonize efficient NHEJ-mediated DSB repair [2,42]. Under normal circumstances where NHEJ rapidly repairs DSBs, the activity of RNF8 and RNF168 in protecting DNA ends from resection may not be required as the broken ends are joined before they can be resected. However, when DNA end joining kinetics is slowed, as may be the case in cells deficient in XLF, which bridges and stabilizes DNA ends, there may be a greater reliance on pathways that protect DNA ends from resection until they can be joined [43][44][45][46].
In response to DNA DSBs, H2AX is phosphorylated to form γH2AX that subsequently promotes the association of RNF8, RNF168 and 53BP1 with chromatin flanking broken DNA ends and these proteins maintain the integrity of DNA ends [5,6,8,11,12,37,[47][48][49]. While neither H2AX nor 53BP1 deficiency alone leads to demonstrable NHEJ defects in lymphocytes, the combined deficiency of H2AX or 53BP1 and XLF leads to an NHEJ defect reminiscent of the NHEJ defect we observed upon depletion of RNF8 or RNF168 in XLF-deficient abl pre-B cells [27,29]. Thus, the NHEJ defect observed in cells with combined deficiencies in XLF and H2AX, 53BP1, RNF8 or RNF168 could reflect delayed joining kinetics (due to XLF deficiency) combined with defective DNA end protection (due to H2AX, 53BP1, RNF8 or RNF168 deficiency).
RNF8 and RNF168 function to recruit 53BP1 to chromatin at DNA DSBs in a manner that depends on their E3 ubiquitin ligase activities [6,8,11,12,15]. Surprisingly, we found that in G 0 /G 1 -phase lymphocytes, the extent of DNA end resection in cells deficient in RNF8 or RNF168 was much less than that observed in cells deficient in 53BP1. Moreover, we find that DNA end resection is diminished in 53BP1-deficient cells that are also deficient in RNF8 or RNF168. This suggests that while RNF8 and RNF168 function to protect DNA ends, they also regulate activities that promote DNA end resection in G 0 /G 1 -phase cells.
Given that RNF8 and RNF168 recruit BRCA1 to DNA DSBs, it is conceivable that they could also recruit proteins that function to promote DNA end resection, such as nucleases, in G 0 /G 1 -phase cells [6,8,11,12]. Moreover, the ubiquitylation of histones in chromatin at DSBs by RNF8 or RNF168 may alter chromatin structure in a manner that makes broken DNA ends more accessible to the resection machinery [4,50].
While DNA end resection is essential for homologous recombination in S/G 2 -phase of the cell cycle, processing of DNA ends is more limited when cells enter the G 0 or G 1 phase of the cell cycle where they rely on NHEJ for efficient DSB repair [2,37,42,51,52]. Extensive DNA end resection in G 0 /G 1 -phase cells would limit NHEJ and promote aberrant homologymediated joining of DNA ends with ssDNA overhangs leading to genome instability [2,42]. However, during NHEJ in G 0 /G 1 -phase cells, some DNA end resection can be required to form compatible DNA ends for joining and for the processing of DNA ends with structures that would otherwise prohibit joining by NHEJ [2,42]. Therefore, in G 0 /G 1 -phase cells, pathways must exist that regulate DNA end resection activities to favor DNA end processing that promotes NHEJ. We propose that RNF8 and RNF168 function to balance pro-and anti-resection activities to provide optimal processing of DNA ends for joining during NHEJ.