Sebastian Veith*, Andrea Schink*, Alexander Bürkle and Aswin Mangerich
Manuscript
* Equal contribution
Abstract
Poly(ADP-ribose) polymerase 1 (PARP1) and WRN are considered as guardians of the genome due to their roles in almost every aspect of DNA metabolism, especially DNA repair and response to genotoxic stress. They have overlapping fields of action, interact with each other and both are localized in the nucleus and also the nucleoli. Here we describe a novel role for PARP1 and poly(ADP-ribosyl)ation in the regulation of WRN’s spatial distribution in the nucleus and the trafficking of WRN out of the nucleoli. We could show that WRN contains four PAR-binding motifs and one of them is located in its RQC domain, which is the binding site for PARP1 and necessary for nucleolar localization. We could demonstrate that PARP1 is essential for efflux of WRN from the nucleoli upon oxidative stress but not for its initial recruitment to, and the containment in, the nucleoli. On the other hand, PARP1 activity and non-covalent PAR-binding to WRN only play a minor role in WRN-trafficking out of the nucleoli, underlining the importance of direct protein-protein interaction between WRN and PARP1. In conclusion, we provide new insight into the role of PARP1 in WRN’s spatio-temporal regulation in response to oxidative stress.
Introduction
The Werner Syndrome protein (WRN) is a member of the RecQL helicase family that comprises five members in humans (RECQL1, WRN, BLM, RECQL4 and RECQL5). So far, little is known about the specific roles or the interplay between the five RecQL helicases, but there is evidence that they can have redundant, synergistic or complementary roles (Croteau et al., 2014). Defects in WRN are the primary cause of the premature aging disorder Werner Syndrome (WS), which is characterized by an early onset of aging-related symptoms such as cataracts, diabetes, gray hair, atherosclerosis, osteoporosis and a higher incidence of cancer (Goto et al., 2013; Lauper et al., 2013). In addition to its helicase and single-strand annealing activities, WRN exhibits, unlike the other RecQL helicases, an exonuclease activity (Rossi et al., 2010a). It is largely agreed upon that WRN is predominantly located in the nucleoli and translocates from the nucleoli upon induction of cellular stress, however, little is known about what drives this process (Constantinou et al., 2000; Gray et al., 1997; Kanagaraj et al., 2012;
Lan et al., 2005; Lee et al., 2005; Singh et al., 2012a). It has been implicated that the AAA ATPase VCP/p97 is involved in this process, however no underlying process has been described so far (Indig et al., 2004; Partridge et al., 2003). Over the last decade, several
publications have shown that acetylation plays an important role in WRN trafficking from and to the nucleolus (Blander et al., 2002; Karmakar and Bohr, 2005; Li et al., 2008; 2010).
Recently, Lee et al. proposed a role of SIRT1 in WRN’s relocalization from the nucleolus, thus strengthening the evidence for a role of acetylation in WRN trafficking (Lee et al., 2015).
Furthermore, phosphorylation reactions appear to contribute to the return of WRN into the nucleoli (Karmakar and Bohr, 2005). WRN has been shown to play a role in almost every aspect of DNA metabolism, including DNA repair, replication, transcription and telomere maintenance (Croteau et al., 2014). Its role in these processes is regulated by protein-protein interactions but also by posttranslational modifications, i.e. phosphorylation, SUMOylation, acetylation or poly(ADP-ribosyl)ation (PARylation) (Adelfalk et al., 2003; Kusumoto et al., 2007; Rossi et al., 2010a). The synthesis of poly(ADP-ribose) (PAR) is catalyzed by poly(ADP-ribose) polymerases (PARP’s) by cleaving NAD+ into ADP-ribose and nicotinamide, with PARP1 being responsible for up to 97% of PAR-production under stress conditions (Shieh et al., 1998). Upon activation PARP1 modifies lysines, glutamates or aspartates at several hundred target proteins, but also heavily automodifies itself with PAR (Gagné et al., 2012; Kawaichi et al., 1981; Ogata et al., 1981). This modification can have various effects: it can in- or decrease enzymatic activity of target proteins, it can alter the spatio-temporal distribution of target proteins and it can lead to complex formation, with PARylated proteins acting as scaffolding factors (Gibson and Kraus, 2012; Mangerich and Bürkle, 2012). Besides covalent modification with PAR, proteins can interact non-covalently with PAR via PAR binding modules (Krietsch et al., 2012). So far, at least 7 different binding modules for non-covalent PAR interaction have been described, with the most prevalent and the least conserved being the PAR binding motif (PBM), first described by Pleschke and colleagues (Altmeyer et al., 2015; Pleschke et al., 2000; Veith and Mangerich, 2014). Proteins bearing a PAR binding module could interact with either PARylated proteins or with free PAR chains, generated by poly(ribose) glycohydrolase (PARG) or ADP-ribosylhydrolase 3 (ARH3) (Niere et al., 2008; Ueda et al., 1972). By now a lot of different roles for PARylation have been described, ranging from participation in spermatogenesis to inflammation, cell death and energy metabolism, but one of its most recognized roles is in DNA repair, where it is implicated in basically every major pathway (Krukenberg et al., 2015;
Mangerich and Bürkle, 2012; Meyer-Ficca et al., 2004). Therefore, it does not come as a surprise that there is increasing evidence of interaction between PARylation and RecQL helicases in general, and with WRN in particular (Veith and Mangerich, 2014). Besides being covalently PARylated, which is still a matter of debate, WRN has been shown to interact with
both PARP1 and PAR and being inhibited by this interaction (Gagné et al., 2012; Khadka et al., 2015; Kobbe et al., 2003a; 2004a; Li et al., 2004; Popp et al., 2012). Also, like WRN, PARP1 is mostly localized in the nucleolus and also its exact functions there remain elusive (Dantzer and Santoro, 2013; Guetg et al., 2012; Meder et al., 2005). The direct interaction between PARP1 and WRN is mediated via PARP’s BRCT and WRN’s RQC domain, the same domain that is required for nucleolar localization of WRN (Kobbe et al., 2003a; 2004a;
Lachapelle et al., 2011; Lebel et al., 2003). However, although there is a bulk of biochemical and some organismal evidence, it is still largely unclear what cellular or organismal consequences these interactions have (Deschênes et al., 2005; Lebel, 2002; Lebel et al., 2003;
Veith and Mangerich, 2014). Here we show that the oxidative stress-induced WRN translocation from the nucleoli is regulated by PARP1 and PAR and that this process can, at least in part, be blocked by PARP inhibitor treatment.
Results
WRN interacts non-covalently with PAR via at least 4 different PBMs
We have previously reported that there are four putative PBM’s in WRN and found that one non-covalently binds PAR strongly and two weakly (Popp et al., 2012). In a follow-up study, we wanted to examine this non-covalent WRN-PAR-interaction further by using several fragments of full length WRN, covering most of its sequence and all but one of the putative PBMs (Figure 33 B, light blue boxes). These fragments were expressed as recombinant proteins and were tested for PAR binding using a far-Western blot assay (PAR overlay).
Unexpectedly, fragment aa 1-520, containing with PBM1 the strongest PBM in our previous screen, displayed only moderate PAR binding under these experimental conditions (Figure 33 A). Moreover, the fragment comprising the PBM that previously showed no PAR binding, revealed a very strong PAR-protein interaction (PBM4, fragment aa 910-1240), closely followed by the fragment containing PBM5, which was not tested before. Using a previously published search pattern we found five PAR-binding motifs in WRN (Figure 33 C) (Popp et al., 2012). Based on these findings we designed several peptides, including the WT sequences of each PBM and several mutated variants thereof (Figure 33 D), and had them synthesized on-membrane (PepSpot Membrane, JPT Technologies). The peptides were then tested for PAR-binding using the PAR overlay assay. In principle, the results of Figure 33 A were confirmed, establishing the putative WRN PBMs 1, 3, 4 and 5 as bona fide PBMs.
Figure 33: WRN has a PAR-binding motif in its nucleolar targeting sequence.
A. PAR overlay western blot of 4 WRN fragments, with BSA and H1 as negative and positive control for PAR-binding, respectively. B. Schematic representation of WRN, with the 5 putative PAR-binding motifs (PBMs) marked in red. RQC, RecQ C-terminal domain; NTS, nucleolar targeting sequence; HRDC, helicase and RNase D-like domain; NLS, nuclear localization signal. The light blue boxes below WRN depict the position and size of the four fragments used in A. C. Pattern that was used to search for PBMs in WRN’s sequence using PattinProt. Blue, basic amino acids (aa); red, hydrophobic aa; X, any aa. Below are the five PBMs in WRN listed that the search yielded. Marked is the position of the core motif by aa number and by bold writing; italic, mismatch compared to search pattern. D. PAR overlay of the PepSpot membrane. Each spot resembles one peptide; the respective sequence is noted to the right. Each PBM found in WRN was tested, along with several mutations. Mutated aa are highlighted in red, essential aa for DNA binding are highlighted in blue. E. 3D structure of the RQC domain of WRN bound to DNA (PDB identifier: 3AAF). DNA is depicted in purple, WRN aa sequence in green, PBM4 in grey, essential aa for DNA binding in blue (Kitano et al., 2010) and crucial aa for PAR binding in red (based on PBM4-mut3). WRN PBM 1: AA 178-185 LKCTETWSLNSLVKHLLGKQ
WRN PBM 2: AA 262-269 NKQLTSISEEVMDLAKHLPH WRN PBM 3: AA 804-811 TYHAGMSFSTRKDIHHRFVR WRN PBM 4: AA 1048-1055 KFMKICALTKKGRNWLHKAN WRN PBM 5: AA 1270-1277 SMAITYSLFQEKKMPLKSIA
N C
By sequentially exchanging several basic aa in the PBM to alanines it was possible to disrupt the PAR-binding capabilities of the respective motif (Figure 33 D). PBM4, which bound strongest in the fragment screen and also strongly in the peptide screen is situated in the one domain of WRN that is necessary for its nucleolar localization, the RQC domain (Figure 33 B, yellow box) (Kobbe and Bohr, 2002). Furthermore, it is also the domain in WRN that is required for interaction with PARP1. We therefore speculated that PARP1, or its product PAR, might be involved in the translocation of WRN out of the nucleoli.
WRN translocates from the nucleoli upon genotoxic stress
Most prior publications reporting the nucleolar localization of WRN failed to use nucleolar proteins as marker to verify the nucleolar localization, thus weakening the results. In order to verify and validate these previous reports, showing that WRN translocates from the nucleoli upon stress induction, we challenged HeLa cells with H2O2 and monitored the time-dependent sub-nuclear relocalization of WRN. We could reproduce findings demonstrating that in unstressed cells WRN indeed localizes mainly in the nucleoli (Figure 34 A&D). After a 10-min treatment however, we observed the first major change in WRN localization: only 40%
of the cells still displayed WRN in the nucleoli and more and more cells had a pan-nuclear distribution of WRN. The second change happened after 30 to 60 min, when suddenly even more WRN left the nucleoli, thus appearing devoid of any WRN (Figure 34 A&C). We classified cells into three different morphological stages during the time line experiment: (i) cells with nucleoli showing a strong WRN signal (full), (ii) cells in which WRN is evenly distributed throughout the nucleus, and (iii) cells with no or weak WRN signal in the nucleoli (empty) (Figure 34 B). Figure 34 D shows that the positive, strong WRN signal, as well as the weak WRN signal, correlates perfectly with the nucleolar marker nucleolin (C23), verifying WRN’s nucleolar localization in unstressed cells. However, the prolonged H2O2 treatment seems to affect the nucleolar architecture, as the nucleolin signal was consistently weaker in treated than in untreated cells (Figure 34 D).
Strangely, sometimes we observed a lot of empty nucleoli even in unstressed cells 1-2 h after treatment. After systematically analyzing this phenomenon, we could demonstrate that the exchange of medium per se, i.e., aspiration of old medium and replacement by fresh medium with or w/o stressor, is a sufficient stressor to trigger WRN’s release from nucleoli (Suppl.
Figure 14, uppermost red frame). We therefore modified our method accordingly, i.e. the medium was not aspirated for treatment but the stressor (or medium as control) was added to
achieve final concentration in the well. All experiments were conducted using this improved method.
Figure 34: WRN relocalizes from nucleolus to nucleoplasm upon oxidative stress.
A. HeLa cells were treated with either 500 µM H2O2 or medium as control and were fixed using 4% PFA after the indicated treatment times. Pictures were taken using confocal microscopy. WRN is depicted in green, Hoechst nuclear staining in blue.
The right-hand WRN pictures are blow-ups of the original picture for better illustration. Shown is one representative experiment out of three independent experiments. B. Exemplary pictures of the different conditions that were analyzed:
nucleoli with strong (red), medium (blue) or no (green) WRN signal. C. Quantification of A, showing mean ± SD. At least 100 cells per experiment were analyzed, n=3. D. WRN resides mainly in nucleoli and leaves upon stress. Nucleolin as nucleolar marker is depicted in red, WRN in green and Hoechst in the merge picture in blue.
Figure 2: WRN relocalizes from nucleolus to nucleoplasm upon
oxida:ve stress
Next, we used camptothecin (CPT) as a stressor that induces a different kind of stress than H2O2, i.e. at submicromolar doses mostly replicative stress and strand-breaks, but also requires WRN for restart of the stalled forks (Thangavel et al., 2015). Since CPT treatment did not elicit such an immediate effect as H2O2 (Figure 34), the cells were treated for up to 6 h and fixed after each hour. When treated with 100 nM CPT, the WRN signal in the nucleoli changed from strong to medium after 1-2 h and most cells had empty nucleoli after 4 or more hours (Figure 35, middle part). With 1 µM CPT however, WRN was released much faster from the nucleoli to the nucleoplasm and the cells displayed empty nucleoli already 1 h after treatment (Figure 35, lowest part). DMSO, which was used as a solvent for CPT, had no effect on WRN’s sub-nuclear localization.
PFA fixation can trigger artificial PAR formation
Even though the methanol:acetic-acid fixation is the method of choice for PAR immunofluorescence, we opted for a 4% PFA fixation method, thus enabling co-staining of WRN and PAR in the same experiment (the WRN antibody is incompatible with the methanol:acetic-acid fixation). In our initial immunofluorescence experiments, we often observed PAR signals in unstressed cells when using the 4% PFA fixation method, even at time point 0 min or in cells treated with PARP inhibitor (Suppl. Figure 15 A, red frame).
Similar findings have previously been reported for ChIP experiments and possible suggested solutions were the use of methanol fixation or the use of PARP inhibitors (Beneke et al., 2012). Since the methanol:acetic-acid fixation was not an option we prevented artificial PARP1 activation during the 4% PFA fixation by adding the PARP inhibitor ABT-888 to the PFA solution, the PBS and the glycine solution. We rationalized that this blocks any artificial PARP activity during the fixation procedure until the cells are lysed with Triton X-100, which leads to a loss of PARP’s substrate NAD+. Indeed, our experiments show that with this modified approach no artificial PAR formation is observed (Suppl. Figure 15 B).
Consequently, all experiments shown here were conducted using this modified protocol.
Furthermore, our experiments revealed a more even PAR signal, when treatment of cells was performed in medium instead of PBS (data not shown). However, since medium inactivates H2O2 partly, the dose was increased accordingly (500 µM H2O2 in medium roughly corresponds to 100 µM in PBS).
Figure 35: Stress-dependent WRN relocalization is not restricted to oxidative stress.
HeLa cells were treated with either 0.1 or 1 µM camptothecin (CPT) or medium as control for the indicated times and were fixed using 4% PFA. WRN is depicted in green, Hoechst nuclear staining in blue. The lower WRN pictures for each treatment condition are blow-ups of the original picture for better illustration. Experiment was conducted once. DMSO control was prepared as control for CPT treatment (dissolved in DMSO).
1h 2h 3h 4h 5h 6h
WRNHoechst Merge
0h
w/o CPT
WRNHoechst Merge 100 nM CPT
WRNHoechst Merge 1 µM CPT
6h
WRNHoechst Merge
3h
0.01 % DMSO
Figure 3: Stress-dependent WRN relocaliza9on is not restricted to
oxida9ve stress.
WRN’s translocation from the nucleoli is almost exclusively PARP1-dependent
The translocation wave of WRN 10 min after H2O2 treatment (Figure 34C) fits very good within the time frame of PAR synthesis, which is strongest between 5 and 10 min after treatment (Martello et al., 2013). To test our hypothesis that PARP1 or PAR regulate WRN’s translocation, we treated HeLa WT cells or HeLa PARP1 KO1 cells (Rank/Veith et al., in preparation (Chapter VI), Figure 36 C) with H2O2 and followed the localization of WRN in the nucleus (Figure 36). Since the most relevant time points seemed to be around 10 min (maximum of PAR formation) and after more than 60 min (complete release of WRN from nucleoli upon stress induction) (Figure 34), we focused on the time points 10 min and 2 h in the following experiments. In untreated cells, we did not observe any difference in the localization of WRN between HeLa WT and PARP1 KO1 cells (Figure 36 A&B). In both cell lines WRN is mostly localized in the nucleoli while some cells show a pan-nuclear distribution (~10%, Figure 36 B, left two columns). This small percentage of cells with pan-nuclear distribution is to be expected since the nucleolar structure is dissolved during cell replication. After a 2-h treatment with H2O2 however, there was a significant difference between the cell lines. While in most WT cells WRN completely disappeared from nucleoli, only few of the PARP1 KO1 cells show no or weak WRN signal in the nucleoli (~30%) and most have WRN still located in the nucleoli (~70%) (Figure 36 A&B, the two right-hand columns).
PARP activity plays a minor role in WRN translocation from the nucleoli
Since the presence of PARP1 is necessary for WRN’s translocation we tested, whether it is through protein-protein interaction, or PARP1’s catalytic activity, or both. To this end, HeLa WT cells were treated with H2O2 and additionally with one of the two different PARP inhibitors ABT-888 (veliparib) or olaparib (Figure 37). Both inhibitors were capable to suppress PARP activity completely as evident from the PAR channel of the immunofluorescence pictures (Figure 37 A). In contrast to the PARP1 KO1 cells however, the inhibitors only partly, but significantly, blocked the stress-induced translocation of WRN from the nucleoli. Whereas almost all untreated cells showed nucleoli with a strong WRN signal, more than half of the H2O2-treated cells displayed empty nucleoli (Figure 37 A&B, the two left-hand columns). Samples treated additionally with PARP inhibitor showed a reduction in cells with WRN-empty nucleoli (~50% reduction) and also displayed some cells with strong WRN signal in the nucleoli, when compared to cells treated with H2O2 only. However,
the effect of the PARP inhibitors was not as strong as observed in PARP1 KO1 cells (Figure 36), suggesting an important role of direct protein-protein interaction between WRN and PARP1.
Figure 36: WRN relocalization is dependent on PARP1.
A. HeLa WT or HeLa PARP1 KO cells were treated with either 500 µM H2O2 or medium as control and were fixed using 4% PFA. WRN is depicted in green, PAR in red and Hoechst nuclear staining in blue. The right-hand WRN pictures are blow-ups of the original picture for better illustration. Shown is one representative experiment out of three independent experiments. B. Quantification of A as mean ± SD. Only cells after two hours treatment were counted. Three independent experiments were analyzed and cells were divided into three categories: with strong, medium or no WRN signal in the nucleoli. At least 100 cells per experiment and treatment were analyzed. p<0.0001 = ****, Chi-squared analysis. C. Cell lysates of HeLa WT and HeLa PARP1 KO cells were separated by SDS-PAGE and blotted on nitrocellulose.
Immunoblotting against PARP1 confirmed that the HeLa PARP1 KO cells are indeed deficient in PARP1.
Figure 4: WRN relocaliza2on is dependent on PARP1.
A)
WRN PAR Hoechst MergeFigure 37: WRN relocalization is partially dependent on PARP activity.
A. HeLa WT cells that received additional PARP inhibitor treatment were pre-treated with either 10 µM olaparib or 10 µM ABT-888 for 30 min. Cells were then treated with either 500 µM H2O2 or medium as control and 10 µM of the respective inhibitor and were fixed using 4% PFA. WRN is depicted in green, PAR in red and Hoechst nuclear staining in blue. The right-hand WRN pictures are blow-ups of the original picture for better illustration. Shown is one representative experiment
A. HeLa WT cells that received additional PARP inhibitor treatment were pre-treated with either 10 µM olaparib or 10 µM ABT-888 for 30 min. Cells were then treated with either 500 µM H2O2 or medium as control and 10 µM of the respective inhibitor and were fixed using 4% PFA. WRN is depicted in green, PAR in red and Hoechst nuclear staining in blue. The right-hand WRN pictures are blow-ups of the original picture for better illustration. Shown is one representative experiment