p38 MAPK signaling and phosphorylations in the BRCT1 domain regulate XRCC1 recruitment to sites of DNA damage

XRCC1 is a scaffold protein involved in base excision repair and single strand break repair. It is a phosphoprotein that contains more than 45 phosphorylation sites, however only a few of these have been characterized and connected to specific kinases and functions. Mitogen activated protein kinases (MAPK) are mediators of cellular stress responses, and here we demonstrate that p38 MAPK signaling is involved in phosphorylation of XRCC1 and regulation of recruitment to oxidative stress. Inhibition of p38 MAPK caused a marked pI shift of XRCC1 towards a less phosphorylated state. Inhibition of p38 also increased the immediate accumulation of XRCC1 at site of DNA damage in a poly(ADP)-ribose (PAR) dependent manner. These results suggest a link between PARylation, p38 signaling and XRCC1 recruitment to DNA damage. Additionally, we characterized two phosphorylation sites, T358 and T367, located within, or close to, the phosphate-binding pocket of XRCC1, which is important for interaction with PAR. Mutation of these sites impairs recruitment of XRCC1 to DNA damage and binding to PARP1/PAR. Collectively, our data suggest that phosphorylation of T358 and T367 and p38 signaling are important for proper regulation of XRCC1 recruitment to DNA damage and thereby avoidance of potential toxic and mutagenic BER-intermediates.

reported phosphorylation sites in XRCC1, but only a fraction of these has been characterized. We examined the phosphorylation state of XRCC1 in response to selected kinase inhibitors, a PARP inhibitor and hydrogen peroxide (H 2 O 2 ) treatment, in order to assess the impact of potential kinases, PARylation and oxidative stress, respectively. Using two-dimensional gel electrophoresis combined with western blot analysis we found that endogenous XRCC1 shifted towards a more alkaline pI when treated with alkaline phosphatase, confirming that XRCC1 is naturally present in a multi-phosphorylated state (Fig. 1a, uncropped images in Supplementary Fig. S1) 26 . Treatment with H 2 O 2 did not substantially change the overall phosphorylation state of XRCC1 (Fig. 1b). Similarly, treatments with a PI3K inhibitor (LY294002) or a PARP inhibitor (4-AN) did not cause any major phosphorylation shifts, i.e. an increase in "tail" (Fig. 1c, quantification of % protein in "tail" and "head" are given in Fig. 1d). In contrast, inhibition of CK2 (TBB) resulted in a considerable shift of XRCC1 towards a less phosphorylated state (Fig. 1c,d). This was expected since XRCC1 contains at least six CK2 phosphorylation sites 20,21 . Interestingly, a clear visible shift towards a less phosphorylated state was also detected after inhibition of p38 (SB203580) (Fig. 1c,d). To verify that p38 is the main target causing these effects, the experiment was repeated using the structurally unrelated p38 inhibitor BIRB0796, as recommended 27 . BIRB0796 treatment also caused a marked shift in pI (Fig. 1c,d), confirming that p38, directly or indirectly via downstream activation of phosphatases and kinases, is involved in regulating the phosphorylation status of XRCC1.
Inhibition of p38 MAPK increases XRCC1 recruitment to sites of DNA damage. The change in phosphorylation status of endogenous XRCC1 after kinase inhibitor treatment prompted us to investigate the effect of these inhibitors on recruitment of a fluorescently tagged XRCC1 construct to DNA damage sites exposed to 405 nm UVA micro-irradiation. In the following experiments, XRCC1-YFP expressing cells were irradiated with a low UV dose (1x dose, see materials and methods) that does not induce DSBs and accumulation of γH2AX, but induces ROS and causes SSBs and base damage 28 . Accumulation of XRCC1 in the irradiated area is shown in Fig. 2a. Inhibition of CK2 or PI3K had no, or minor, effects on XRCC1 accumulation at micro-irradiation foci (mIF) (Fig. 2b,c). Interestingly, inhibition of p38 by both SB203580 and BIRB0796, markedly increased XRCC1 accumulation at mIF (Fig. 2d,e). This suggests a role of p38 in the regulation of XRCC1 recruitment to ROS induced DNA damage.
To narrow down the regions for possible phosphorylation sites affected by p38 signaling, we tested the response of p38 inhibition on two deletion mutants of XRCC1, MD and BLB, both which contain the BRCT1 domain (Fig. 2f) and are recruited to mIF 29 . The MD mutant and full-length XRCC1 had a similar increase in recruitment to mIF, while the BLB mutant displayed a reduced accumulation. Still, an increase in accumulation after p38 inhibition was observed also for BLB (Fig. 2g). This suggests that phosphorylation sites in the BRCT1 region, that is common for both MD and BLB, might be directly or indirectly affected by p38 signaling. In order to avoid the many phosphorylation sites in the inter-BRCT1 domain, we used the MD deletion mutant for the following experiments. p38 signaling regulates the kinetics of repair and is dependent on PARP1 activity. To assess whether the accumulation of XRCC1 at sites of DNA damage after p38 inhibition affected XRCC1's ability to . Recombinant IRF3 (Interferon regulatory factor 3) was added to all the lysates to serve as an internal positional reference during picture alignment (dotted lines). IRF3 is depicted only in the 2D-PAGE western analysis of BIRB0796 treated cells for simplicity. Cropped images of one out of two gels with similar patterns is depicted. (d) Quantification of %Head and %Tail (defined in C) using the Kodak Molecular Imaging software version 4.0.1. Ratios of "head" and "tail" densities versus total XRCC1 intensity (head + tail) were calculated after subtracting background intensity values. support BER/SSBR, we performed Comet assays on cells treated with inhibitors and H 2 O 2 . Inhibition of PI3Ks (LY294002), used as a negative control in this setting as it did not affect recruitment of XRCC1, did not alter the average genomic fragmentation compared to mock treated cells (Fig. 3a). p38 inhibition (SB203580 and BIRB0796), on the other hand, resulted in a significant increase in initial genomic fragmentation. This was also observed for inhibition of PARylation (4-AN) and CK2 (TBB) (Fig. 3a), in accordance with previous reports 20,21,30 . No increase in genomic fragmentation compared to mock treated controls was observed in inhibitor treated cells in absence of H 2 O 2 (data not shown). After 20 min recovery, only the cells treated with PARP inhibitor (4-AN) had increased levels of strand breaks compared to control (Fig. 3b). This suggests that increased XRCC1 accumulation leads to increased levels of AP sites or strand breaks in regions of oxidative DNA damage. Lesions and repair-complex-DNA intermediates that occur in close proximity in different strands will appear as strand breaks in the Comet assay. Repair intermediates such as AP sites and strand breaks are potentially toxic and mutagenic, thus tight regulation of BER/SSBR is important. Notably, increased accumulation of XRCC1 at sites of DNA damage and increased transient DNA fragmentation after inhibition of p38 may represent undesired effects that should be considered when using p38 inhibitors as anti-inflammatory drugs.
Efficient XRCC1 recruitment to UVA induced damage sites is dependent on PARP1 and PARylation 2, 3 , thus we examined whether the increased recruitment after p38 inhibition was dependent on PARylation. We found that inhibition of PARP1 with PJ34, which does not trap PARP1 to site of DNA damage 28 , abolished the recruitment of MD to mIF. Furthermore, inhibition of p38 did not increase the recruitment in absence of PARylation (Fig. 3c). In support of a link between PARylation and activation of p38 signaling, it has been reported that p38 activation is impaired in PARP1 deficient cells 31,32 . Additionally, PARP1 activity by itself increases ROS production, which may contribute to amplification of ROS mediated p38 activation 33,34 .
In summary, our data demonstrate that XRCC1 recruitment to mIF is dependent on PARylation and modulated by p38 signaling. However, the exact connection between p38 signaling and PARylation remains elusive. We hypothesize that PARylation at SSBs is involved in activation of p38. Our data demonstrate that p38 mediated signaling regulates the phosphorylation status of XRCC1, and thereby its recruitment to DNA damage. When phosphorylation of XRCC1 is reduced, more XRCC1 is recruited to sites of DNA damage, leading to a transient increase in AP sites and strand breaks (Model illustrated in Fig. 3d).
Mutations in XRCC1 phosphorylation sites in the BRCT1 domain impair recruitment to DNA damage. We performed a search for predicted p38 as well as general MAPK phosphorylation sites in XRCC1 in silico using five different prediction algorithms. From the predicted sites (see Fig. 4a), we selected putative phosphorylation sites for serine/threonine (S/T) to alanine (A) mutagenesis based on the following criteria: 1) suggested by more than two prediction algorithms 2) reported to be phosphorylated 3) within MD (aa 166-436). The candidate phosphorylation sites fulfilling these criteria were S226, T257 and S266 (Fig. 4a). Considering the importance of the BRCTI domain in recruitment of XRCC1 to sites of DNA damage 5, 29 we also included T367. This is the only predicted MAPK site within the BRCT1 domain and it was recently shown to be phosphorylated 35 and part of the phosphate-binding pocket 5 . However, the functional role of phosphorylation at T367 is not known.
Micro-irradiation experiments showed that the S226A, T257A, and S266A mutants responded to p38 inhibition similarly to wild type ( Supplementary Fig. S2). Phosphorylations at T257 and S266 have recently been suggested to be involved in nuclear import and binding to Importin α 36 , however we did not observe any change in nuclear localization of the single mutants. Mutation of T367 on the other hand, almost completely abrogated recruitment of MD to mIF. Inhibition of p38 induced a low and transient increase in recruitment (Fig. 4b), suggesting that T367 could be one, but not the only target of p38 signaling. Similar results were obtained for full-length XRCC1 T367A (data not shown). XRCC1 deficient cells complemented with full-length XRCC1 T367A accumulated almost as much genomic fragmentation as XRCC1 deficient cells after oxidative stress (Fig. 4c). XRCC1 T367A cells had an increased repair rate compared to XRCC1 deficient cells, but a slower repair rate than wild type XRCC1 complemented cells (Fig. 4c). This indicates that the attenuated recruitment of XRCC1 T367A to DNA damage sites caused a reduced rate of DNA repair.
Next, we analyzed XRCC1 isolated from cells treated with H 2 O 2 and/or p38 inhibitor by mass spectrometry (MS), in order to detect possible dynamic changes of T367 phosphorylation in response p38 inhibition and oxidative stress. We were able to identify phosphorylated forms of the peptide containing T367 (Fig. 4d). However, the phosphorylations of this peptide were assigned to two other sites, S357 and T358, not previously reported to be phosphorylated ( Supplementary Fig. S3). We could not determine changes in phosphorylation of these sites upon p38 inhibition and oxidative stress, possibly due to low levels of the phosphorylated peptide.
The three phosphorylation sites S357, T358 and T367, are all within or close to the phosphate-binding pocket of XRCC1 important for interaction with PAR 4, 5 (Fig. 4d). Mutations of R335 and K369 were recently shown to impair XRCC1 interaction with PAR, and thereby recruitment of XRCC1 to sites of DNA damage 5 . Therefore, we next investigated whether mutations of the two novel phosphorylation sites identified, S357 and T358, affected recruitment of MD to mIF. The mutant S357A was similarly recruited as wild type MD, while the T358A mutation abolished accumulation in mIF (Fig. 4e,f). Inhibition of p38 did not induce any accumulation of T358A (Fig. 4f). Using four different algorithms for prediction of kinase consensus sites we found that all three sites, S357, T358 and T367, could be phosphorylated by MAPKs or substrates of p38 MAPK (Table 1). Among several other potential kinases, one algorithm proposed the central DDR kinase DNA-PK as a candidate for all three residues. Together, these results suggest that the T358 residue, in addition to T367, could be targets of p38 signaling.
To explore the function of the two phosphorylation sites T358 and T367 further, we created phospho-mimetic aspartate (D) mutants and analyzed recruitment to mIF. Both A and D mutants had strongly reduced ability to accumulate at site of DNA damage compared to wild type even at a high laser dose (5x) (Fig. 5a,b). However, T358D was clearly recruited whereas minor accumulation was observed for T367A. Inhibition of p38 increased recruitment of T358D similarly to T367A (data not shown and Fig. 4b), suggesting that none of these are the only target of p38 signaling. Although all mutants are localized to nuclei, they contained variable amounts of nuclear foci. Wild type MD appeared in nuclear foci in most cells (Fig. 5c). These foci represent both stress induced foci and replication foci 29 . In contrast, the MD mutants did not appear in nuclear foci, with exception of T358D that localized to nuclear foci in some cells, but at a reduced strength compared to wild type (Fig. 5c). All MD wild type foci, also those colocalizing with PCNA and thus representing replication foci, disappeared after treatment with PARP inhibitor (Fig. 5d). When we examined the MD phospho-mutants' ability to colocalize with PCNA in replication foci they all colocalized to some extent in XRCC1 deficient cells (CHO EM9), but less or not at all in HeLa cells (Supplementary Fig. S4). This could indicate that the mutants are to some degree outcompeted by endogenous XRCC1 in HeLa cells, which could be explained by reduced affinity for PARylated proteins. Hence, we next examined the PAR-and/or PARP1-binding properties of the MD mutants by immunoprecipitations in presence and absence of two PARP inhibitors. Immunoblotting revealed that higher levels of PARP1 were pulled down with wild type MD compared to all mutants. T367A pulled down some PARP1, while no detectable PARP1 was pulled down with T367D and T358A/D (Fig. 5e, Supplementary Fig. S5a,b). We also examined the levels of the PAR dependent E3 ubiquitin ligase RNF146, known to target PARP1 and several other repair proteins including XRCC1 for degradation 8 . The levels of RNF146 in the pull down followed the same pattern as PARP1 (Fig. 5e, Supplementary Fig. S5a,c). Interestingly, wild type MD pulled down more PARP1 and RNF146 from cells treated with PARP inhibitor (Fig. 5e, Supplementary Fig. S5b,c). This was observed using both PJ34 and 4-AN (data not shown) at doses that abolish nuclear PARylation 28 . These results deviate from previously published pull down experiments using a MD like domain of XRCC1 (aa 161-406) and the PARP inhibitor KU58948 5 . The reason for this is not known, but it could be due to differences in the PARP inhibitors ability to interfere with PARP1 binding to chromatin. PARP1 trapping can be a problem in such studies and has been suggested for several PARP inhibitors (4-AN, niraparib, olaparib and veliparib) 28,37 .
To further examine the PAR-binding properties of the MD mutants we washed immunoprecipitates in high salt buffer (1 M) to remove low affinity binders. All mutants pulled down less PARylated proteins compared to wild type, only a few weak bands were detected in the T367A and T358D pull downs (Fig. 5f, quantification in Supplementary Fig. S5d). These data support that phosphorylation of these residues are important for regulating the affinity to PARylated proteins.
In summary, the results presented here demonstrate a role of p38 signaling and phosphorylations in the BRCT1 domain of XRCC1 for accumulation at sites of DNA damage as well as sites of replication. The p38 signaling pathway has been extensively studied as mediator of cell cycle regulation and apoptosis in response to ROS, UV-irradiation, and/or DNA-lesions (reviewed in ref. 38), but this is to our knowledge the first time it has been shown to influence the dynamics of recruitment of DNA repair proteins. 2D-PAGE analysis demonstrated that p38 clearly alters the phosphorylation status of XRCC1. However, XRCC1 is not necessarily a direct substrate of p38. We show that T367 in XRCC1, an in silico predicted p38 phosphorylation site, is a functional phosphorylation site that has importance for binding to PAR and thereby recruitment to sites of DNA damage. We also identify T358 as a novel functional phosphorylation residue that affects the affinity to PAR. These results demonstrate that phosphorylations in or around the phosphate-binding pocket are important for a dynamic regulation of affinity to PAR and thereby recruitment to DNA damage.

Preparation of cell lysates. Exponentially growing
Alkaline phosphatase treatment. Endogenous XRCC1 from HeLa cells was immunoprecipitated over night at 4 °C using Dynabeads protein A magnetic beads coupled to monoclonal XRCC1 antibodies (ab1838, Abcam). The beads were washed three times with a 10 mM Tris-HCl pH 8 and 50 mM KCl buffer, and divided in two equal parts prior to resuspension in phosphatase buffer (10 mM Tris-HCl pH 7.9, 50 mM NaCl, 10 mM MgCl 2 , 1 mM DTT, and 0.1 mM ZnCl 2 ). Excess of calf intestinal phosphatase (CIP, Biolabs) was added to one of the tubes following incubation for 1 hour at 37 °C. Proteins were eluted from beads by overnight incubation in Destreak solution containing 1% IPG buffer pH 4-7 (GE Healthcare). Eluates were collected and submitted to 2D-PAGE as described above. Comet assay (alkaline single cell gel electrophoresis). HeLa S3 were pre-treated for 1 hour at 37 °C with 0.25% DMSO (mock), 10 µM 4-amino-1,8-naphthaliamid (4-AN), or 25 µM of a kinase inhibitor; TBB, LY294002, SB203580, BIRB0796. These cells and CHO EM9 stably transfected with XRCC1-YFP T367A, XRCC1-YFP or YFP were next exposed to 10 min of 47 or 62 µM H 2 O 2 (Sigma-Aldrich) in PBS at 37 °C. The cells were washed twice with PBS, harvest or kept in growth medium with the respective inhibitors for further recovery. Cells were harvested in ice cold 30% FCS in PBS and subjected to lysis ON, alkaline DNA unwinding (pH > 13.3), and single cell electrophoresis as described 28 . 100 comets were selected randomly from each slide and evaluated using Komet 5.0 imaging software (Andor Technology).

XRCC1 map and predictions. Predictions of interaction motifs and p38 phosphorylation sites on XRCC1
were performed by using several different online services. XRCC1 protein sequence was uploaded as FASTA format acquired from UniProt 41 , as reference to its UniProt identifier (Uniprot ID: p18887). If possible all isoforms of p38 were tested, if not the prediction was narrowed down to p38 α (also known as MAPK14 . Whole cell extracts were prepared as described above, with exception of PJ34, that were added to buffer I to a concentration of 10 µM prior to lysis. Beads were washed three times in 10 mM Tris-HCl pH 7.5 containing either 100 mM or 1 M NaCl. Elution: Immunoprecipitated proteins were eluted in LDS loading buffer (Invitrogen) containing 100 mM DTT by heating the beads for 10 minutes at 70 °C and separated on a NuPAGE 4-12% Bis-Tris protein gel (Invitrogen) using MOPS SDS running buffer (Invitrogen). Proteins were blotted to polyvinylidene fluoride (PVDF) membranes (Immobilon, Millipore).
Mass spectrometry (MS) analysis. Bands on the M.W. area of XRCC1 were excised and submitted to in-gel tryptic digestion. Tryptic digests were dried out, resuspended in 0.1% formic acid and analysed on a Thermo Scientific Orbitrap Elite mass spectrometer coupled to an Easy-nLC 1000 UHPLC system (Thermo Scientific). Peptides were injected onto a Acclaim PepMap100 C-18 column (75 μm i.d. × 2 cm, C18, 5 μm, 100 Å) (Thermo Scientific) and further separated on a Acclaim PepMap RSLC Nanoviper C-18 analytical column (50 μm i.d. × 15 cm, C18, 2 μm, 100 Å) (Thermo Scientific). An 80-minute method with a 55 minutes' gradient starting with 100% Buffer A (0.1% Formic acid) with an increase in buffer B (100% Acetonitrile, 0.1% Formic acid) until 30% and a 250 nL/minute flow rate was employed. The peptides eluting from the column were analyzed in positive-ion mode under data dependent neutral loss MS3 mode using CID fragmentation with normalized collision energy 35. Each MS scan (m/z 400-1600) was acquired at a resolution of 120,000 FWHM, followed by 10 MS/MS scans triggered for intensities above 1.7E4. MS spectra were analyzed using Thermo Proteome Discoverer version 1.4.0.288 software running Mascot and the Sequest HT search algorithms. Spectra were searched against a Human RefSeq database with the following parameters: Max. missed cleavage = 2, precursor mass tolerance = 10 ppm, fragment mass tolerance = 0.05 Da, static modification: carbamidomethyl (C: +57.021 Da), dynamic modifications: Phospho (S,T,Y: +79.966), Oxidation (M: +15.995). Proteome Discoverer assign peptides identified with high degree of confidence as having False Discovery Rate (FDR) ≤ 0.01, medium confidence FDR ≤ 0.05, and low peptide confidence for FDR lower than 5%. The PhosphoRS 3.0 algorithm was also used to assign phosphorylations sites in XRCC1 and to calculate a site probability score, which determines confidence in phosphorylation site localization.