Dynamic protein interaction networks such as DNA double-strand break (DSB) signaling are modulated by post-translational modifications. The DNA repair factor 53BP1 is a rare example of a protein whose post-translational modification-binding function can be switched on and off. 53BP1 is recruited to DSBs by recognizing histone lysine methylation within chromatin, an activity directly inhibited by the 53BP1-binding protein TIRR. X-ray crystal structures of TIRR and a designer protein bound to 53BP1 now reveal a unique regulatory mechanism in which an intricate binding area centered on an essential TIRR arginine residue blocks the methylated-chromatin-binding surface of 53BP1. A 53BP1 separation-of-function mutation that abolishes TIRR-mediated regulation in cells renders 53BP1 hyperactive in response to DSBs, highlighting the key inhibitory function of TIRR. This 53BP1 inhibition is relieved by TIRR-interacting RNA molecules, providing proof-of-principle of RNA-triggered 53BP1 recruitment to DSBs.
Post-translational modifications (PTMs) control the functional assembly of numerous protein complexes. In the cascade of PTMs triggered by DNA DSBs in mammals, the DNA damage response (DDR) p53-binding protein 1 (53BP1) is recruited to damaged chromatin by recognizing histone H2A ubiquitylated at Lys15 (H2AK15ub) and histone H4 dimethylated at Lys20 (H4K20me2) in the nucleosome core particle (NCP-ubme)1,2,3,4,5. 53BP1 plays an important role in maintaining the balance between the nonhomologous end joining (NHEJ) and homology-dependent DNA repair pathways6,7,8. 53BP1 favors NHEJ over homology-dependent repair (HDR) by inactivating DNA end resection, the initiation step of HDR, and by blocking the recruitment of HDR factor BRCA1 to DSBs9,10. Loss or inhibition of 53BP1 promotes HDR7,8,11. As an activator of NHEJ, 53BP1 also promotes immunoglobulin class switch recombination (CSR)12,13. 53BP1 represents a rare example of a protein whose PTM reader function can be inactivated. While 53BP1, via its tandem Tudor domain, recognizes the constitutive PTM H4K20me2 in damaged chromatin, in the absence of damage, the ability of 53BP1 to interact with H4K20me2 is inhibited by the protein TIRR14. The mechanism for the inhibitory function of TIRR is not known, and its proposed mode of action has been controversial14,15. How 53BP1 dissociates from TIRR in response to DNA damage is also unknown.
Here we show that TIRR directly inhibits the interaction of 53BP1 with NCP-ubme. An X-ray structure of TIRR–53BP1 reveals an intricate binding area, centered on an arginine residue in TIRR, that blocks the histone binding surface of the 53BP1 tandem Tudor domain. This unique binding mechanism is highly specific for 53BP1, as shown by mass spectrometry and mutagenesis. Based on the TIRR–53BP1 structure, we designed a separation-of-function 53BP1 mutant inactive for binding TIRR but fully functional for DSB recruitment. The ‘hyperactive’ nature of this 53BP1 mutant demonstrates that a major function of TIRR is to keep 53BP1 in an inactive state in the absence of DNA damage. We also address the mechanism of 53BP1 dissociation from TIRR. As TIRR is an RNA-binding protein16,17 and noncoding RNAs have been implicated in the recruitment of 53BP1 to DSBs18,19,20, we examined the possibility that RNA molecules produced in response to DNA damage could disassemble the TIRR–53BP1 complex. We engineered the TIRR-related nucleotide- and RNA-binding and processing enzyme NUDT16 into a 53BP1-binding protein (NUTD16TI) and validated the protein design using X-ray structure determination of NUDT16TI–53BP1 and quantitative binding assays. Nucleotides dissociated 53BP1 from NUDT16TI, which led us to show that RNA molecules also disassemble the TIRR–53BP1 complex. RNA molecules may therefore serve as a trigger for 53BP1 chromatin recruitment in response to DNA damage.
TIRR blocks the association of 53BP1 with the modified nucleosome core particle
Interaction of TIRR with the Tudor domains of 53BP1 (53BP1-Tudor) does not require lysine or arginine methylation and has a higher binding affinity than typical for PTM reader domains14. As we previously showed that TIRR overexpression in mammalian cells abolished formation of 53BP1 ionizing radiation-induced foci (IRIF)14, we examined whether TIRR inhibits the interaction of 53BP1 with its minimal chromatin substrate, NCP-ubme1,2. While a GST-fused 53BP1 fragment encompassing the Tudor domains and ubiquitin-dependent recognition motif readily interacted with NCP-ubme, as previously reported2,3,21, the same 53BP1 construct bound to TIRR had no affinity for NCP-ubme. Therefore, TIRR most likely directly blocks 53BP1 recruitment to chromatin (Fig. 1a).
The X-ray structure of TIRR–53BP1 reveals a binding interface centered on Arg107. To understand how TIRR inhibits 53BP1 binding to chromatin, we determined the X-ray structure of human TIRR in complex with 53BP1-Tudor (hereafter referred to as TIRR–53BP1), at 2.18 Å resolution (Table 1). In the crystal asymmetric unit there are two TIRR homodimers, each interacting with one 53BP1 molecule (Supplementary Fig. 1a). As dictated by symmetry, in solution each TIRR protomer should bind one 53BP1 molecule (Fig. 1b). Noticeably, the N termini of the two 53BP1-Tudor molecules point away from the TIRR dimer and toward the previously identified upstream oligomerization region of 53BP122. Oligomerization of 53BP1 is essential for its function, but its precise oligomerization state had not been established when we began this study2,23. We have now reconstituted a stable 400-residue 53BP1 construct, comprising the oligomerization region and Tudor domains, and sedimentation-velocity analytical ultracentrifugation revealed that 53BP1 is a homodimer and thus expected to ‘clamp’ a TIRR homodimer (Fig. 1c).
In the X-ray structure, TIRR masks the H4K20me2 binding surface of 53BP1, but its binding mode differs radically from that of H4K20me4. It involves a ~640-Å2 binding surface that includes the two Tudor domains (Fig. 1d). In the H4K20me2–53BP1 complex, the binding surface area is smaller (~257 Å2) and interaction is mainly mediated by the first Tudor domain and a few residues in the inter-Tudor region1,24 (Fig. 1e). A binding cage formed by four aromatic residues and Asp1521 in the first Tudor domain accommodates the dimethyl-lysine in H4K20me2 via cation-π interactions and an ion pair with the dimethylammonium group of K20me2.
In TIRR, a loop comprised of residues 101 to 107 connecting two antiparallel β-strands is the main interface with 53BP1 (Figs. 1d and 2a). TIRR Arg107 is central to this binding site and occupies a unique intermolecular cavity assembled from TIRR Trp24, and 53BP1 Tyr1502, Asp1521 and Met1584 (Fig. 2a,c). The guanidinium group of Arg107 forms hydrogen bonds with the carboxyl group of Asp1521 and carbonyl group of Met1584. TIRR Gly21 amide group and Trp24 side chain HE1 atom are hydrogen-bonded to the carbonyl and carboxyl groups of Asp1521, respectively. TIRR Pro105 contributes van der Waals contacts and CH–π interactions, with its side chain sandwiched between the aromatic rings of 53BP1 Tyr1500 and Phe1553 (Fig. 2a,d). The carbonyl group of Pro105 forms a hydrogen bond with the hydroxyl group of Tyr1502, and the TIRR His106 HE2 atom may be transiently hydrogen-bonded to the backbone carbonyl group of 53BP1 Glu1551. TIRR residues 102 to 104 do not interact with 53BP1, but Leu101 contacts the side chain of 53BP1 Trp1495. Distant from the aforementioned binding loop, TIRR Leu20 contacts Tyr1523 while TIRR Lys10 participates in van der Waals interaction with the aromatic ring of Trp1495. Additionally, the ammonium group of Lys10 is hydrogen-bonded to the carbonyl group of Trp1495 and to the hydroxyl group of Tyr1523 (Fig. 2b,e). In the other protomer, the ammonium group of TIRR Lys151 is positioned to form salt bridges with the carboxyl groups of evolutionarily conserved 53BP1 residues Glu1549 and Glu1551, an observation consistent with NMR chemical shift perturbations indicating that Lys151 is in the vicinity of the binding interface14 (Fig. 1d). However, these salt bridges are not essential for complex formation, as the K151E mutation does not disrupt the interaction of 53BP1 with TIRR in cells14.
53BP1 and TIRR undergo conformational changes upon complex formation
Substantial conformational changes in 53BP1 result from its interaction with TIRR. The 53BP1 loop (residues 1,495–1,499) harboring Trp1495, a residue essential for methyl-lysine recognition, is flipped by about 180°, and the Trp1495 aromatic ring is reoriented to interact with TIRR Lys10 and Leu101 (Figs. 1d–f and 2b,e). Tyr1523 in the methyl-lysine binding cage of 53BP1 also undergoes a drastic reorientation in complex with TIRR, as does Phe1553 and the loop (residues 1,547–1,553) harboring this residue (Figs. 1d–f and 2b,e). Noticeably, the side chains of Trp1495 and Tyr1523 in apo-53BP1 are highly flexible, reflecting the structural malleability of 53BP11,25. There are also changes in TIRR upon interaction with 53BP1. In the X-ray structures of apo-TIRR and TIRR protomers not interacting with 53BP1 within TIRR–53BP1 crystals, Lys10 and residues 103–107 of the 53BP1-binding loop are highly flexible, as deduced from the poor electron density for these regions. The same residues are more rigid in the TIRR–53BP1 complex, providing well-defined electron density. The selective broadening of NMR spectroscopy signals we previously observed for 13C-methyl-labeled Lys10 upon TIRR–53BP1 complex formation further highlights the reduced flexibility of this key interface residue14. The agreement between the binding interface mapped by NMR spectroscopy14 and the intermolecular contacts in the X-ray structure of TIRR–53BP1 suggests these contacts are not artifacts of crystallization.
The functional effects of mutations in TIRR and 53BP1 validate the 3D structure of TIRR–53BP1
Although not an enzyme, TIRR is evolutionarily related to the Nudix hydrolase NUDT16 (47.9% amino acid sequence identity and 60.5% homology)16,26,27, an RNA nucleotide diphosphatase that processes the 5′ m7GpppG cap from small nucleolar RNAs and cytoplasmic messenger RNAs28. In addition, NUDT16 hydrolyzes inosine diphosphate into inosine monophospate (IMP) and inorganic phosphate, thereby limiting inosine incorporation in RNA27,29,30, and also processes ADP-ribosylation of proteins31,32. Notably, although NUTD16 does not bind 53BP1 (see below), the TIRR loop residues mediating interaction with 53BP1 are conserved in NUDT16 (Fig. 3a). A major difference between TIRR and NUDT16, however, is the lack of a histidine (TIRR His106) between Pro104 and Arg105 in NUDT16, corresponding to Pro105 and Arg107 in TIRR (Fig. 3a). As a result, the loop in NUDT16 is shorter and rigid, and locked in a conformation incompatible for 53BP1 binding (Supplementary Fig. 2).
Guided by the TIRR–53BP1 structure and the structural similarity of TIRR and NUDT16, we assessed the effect of 53BP1 and TIRR mutations on the TIRR–53BP1 complex formation in vitro using proteins expressed in Escherichia coli and in mammalian cells (Fig. 3b). Mutating interfacial residues in 53BP1 diminished its interaction with TIRR. Replacing residues 101 to 107 in the TIRR 53BP1-binding loop with the corresponding residues in NUDT16 abolished formation of the TIRR–53BP1 complex. Single-point TIRR mutations K10E and R107S likewise abolished 53BP1 binding, while H106G, H106E, and deletion of H106 (H106Δ) greatly diminished the interaction. Furthermore, mass spectrometric analysis of proteins associated with wild-type (WT) TIRR and TIRR variants containing the loop residues of NUDT16 demonstrated that the TIRR loop motif is highly specific for 53BP1 interaction (Supplementary Fig. 3 and Supplementary Table 1). The K151E mutation, which we expected to disrupt salt bridges between TIRR and 53BP1 (Fig. 1d), did not affect their interaction. We previously observed that TIRR overexpression in mammalian cells blocked 53BP1 recruitment to DSBs, thereby impairing 53BP1 function in the DDR14. Unlike WT TIRR, expression of the K10E and R107S mutants in RPE1 cells did not prevent 53BP1 IRIF formation (Fig. 3c).
A separation-of-function mutation creates a hyperactive form of 53BP1
Residues Trp1495 and Asp1521 in the first Tudor domain of 53BP1 participate in critical intermolecular contacts in the TIRR–53BP1 and H4K20me2–53BP1 structures (Fig. 2). W1495A and D1521A mutations markedly diminished 53BP1 binding to TIRR (Fig. 3b) and H4K20me21,14, and they abolished 53BP1 IRIF formation1,14. To create a version of 53BP1 that does not bind TIRR but could still be recruited to DNA damage sites, we mutated Phe1553 in the second Tudor domain. This residue is rigid and buried at the intermolecular interface in the TIRR–53BP1 structure (Fig. 4a), but is flexible in the 53BP1–H4K20me2 structure (Fig. 1e,f). In one conformation, the phenyl ring points away from H4K20me2, suggesting that Phe1553 is not essential for the 53BP1–H4K20me2 interaction. Indeed, mutating Phe1553 to arginine prevented the interaction of 53BP1 Tudor domains with TIRR (Fig. 4b). Formation of intense IRIF by the minimal focus-forming region (FFR, residues 1,220–1,711) of 53BP1 F1553R, demonstrating interaction with H4K20me2, and the markedly diminished interaction of F1553R with TIRR in immunoprecipitation assays validated this separation-of-function mutant in mammalian cells (Fig. 4b). As negative controls, we mutated two solvent-accessible residues, Glu1551 and Tyr1552, adjacent to Phe1553. As predicted from the TIRR–53BP1 structure, E1551R and Y1552R mutations had virtually no effect on the interaction of 53BP1 with TIRR in cells.
We previously observed that loss of TIRR markedly reduced the soluble levels of 53BP1 in nuclei and enhanced phosphorylation and association of 53BP1 with its effector proteins after DNA damage14. Notably, 53BP1 F1553R mimics the loss of TIRR, giving rise to a hyperactive form of 53BP1. Salt fractionation assays with F1553R revealed a distinct decrease in nuclear soluble 53BP1 (Fig. 4c). In response to ionizing radiation, 53BP1 F1553R was efficiently phosphorylated at Ser25 and Ser29, and it associated more strongly than WT 53BP1 with known 53BP1 effector proteins such as PAX transactivation domain-interacting protein (PTIP) and topoisomerase IIβ binding protein 1 (TopBP1; Fig. 4d). This hyperactivity of 53BP1 F1553R correlated with a substantial increase in mobility in cells compared to WT 53BP1, as shown by fluorescence recovery after photobleaching (FRAP) assays (Fig. 4e). The value of halftime recovery (t½) was 55.6 s for cells expressing WT 53BP1, which is more than 3 × slower than for the cells expressing 53BP1 F1553R (15.5 s). This increased mobility of 53BP1 F1553R mutant may be the basis of its hyperactive state.
We next checked whether, similarly to loss of TIRR or ectopic expression of the histone H2A Lys15-specific E3 ubiquitin ligase RNF16814,33, 53BP1 F1553R would impact DNA repair. We examined the influence of F1553R on the cell-killing effect of a poly-(ADP-ribose) polymerase inhibitor (PARPi) on DNA repair-deficient cells. We previously reported that loss of 53BP1 function due to TIRR overexpression conferred resistance to the PARPi olaparib via reactivation of the HDR pathway in mouse embryonic fibroblasts (MEFs) carrying hypomorphic BRCA1-Δ11 alleles14. In contrast, TIRR depletion or RNF168 overexpression increased the olaparib sensitivity of BRCA1-mutant MEFs14,33. To determine whether the F1553R 53BP1 mutation sensitized cells to olaparib, we reconstituted 53BP1−/− MEFs with WT or F1553R human 53BP1 (Fig. 5a). As anticipated, BRCA1 depletion in the WT-expressing MEFs made the cells more sensitive to olaparib (Fig. 5b). Expression of F1553R in the context of BRCA1 depletion further sensitized these cells to olaparib, thereby phenocopying the TIRR-deficient phenotype (Fig. 5c). Consistent with increased inactivation of HDR, the number of RAD51 foci per cell was smaller in 53BP1Δ RPE1 cells reconstituted with F1553R relative to WT 53BP1 (Fig. 5d,e). Like the TIRR-deficient cells or cells expressing TIRR K10E14, a mutant that does not bind 53BP1, F1553R-expressing 53BP1−/− MEFs were defective in DNA damage repair. There was increased persistence of DSBs, as gauged from the kinetics of histone γH2AX IRIF formation (Fig. 5f).
Binding of 53BP1 to damaged chromosome ends is necessary for fusion of uncapped telomeres34, and thus 53BP1 activity at telomeres can be quantified by telomeric fluorescence in situ hybridization (FISH) assays. Errors in telomeric replication35 or choice of DSB repair pathways (classical versus alternative NHEJ) lead to telomere loss in fused chromosomes36. Therefore, we scored chromosome fusions by monitoring the FISH signal in cells expressing the dominant-negative TRF2ΔBΔM allele (Fig. 5g), which promotes NHEJ at chromosome ends37. We transiently expressed TRF2ΔBΔM in 53BP1Δ RPE1 cells reconstituted with WT 53BP1 or F1553R mutant, and metaphase spreads were prepared and scored for chromosome end-to-end fusions. 53BP1 deficiency resulted in diminished fusion events at telomeres induced by transient TRF2ΔBΔM expression, and the total numbers of chromosome fusions were comparable in WT and F1553R 53BP1 (Fig. 5g). Notably, the F1553R mutant led to a substantial increase in telomere-free fusions, suggesting that TIRR–53BP1 interaction prevents the loss of uncapped telomeres before NHEJ-mediated fusion. The underlying mechanism for this phenotype remains unclear, but may involve telomeric replication, which would be disrupted by the F1553R 53BP1 mutant.
To examine the function of TIRR–53BP1 interactions in regulating NHEJ, we monitored the effect of the 53BP1 F1553R on CSR in cultures of primary splenic B cells from 53BP1−/− mice. Class switching to immunoglobulin G1 (IgG1) was analyzed in 53BP1−/− B cells stimulated with anti-CD40 antibody and IL-4 upon reconstitution with retroviruses expressing WT 53BP1, the F1553R mutant, or the H4K20me2-binding-defective W1495A mutant (Fig. 5h). Note that all rescue constructs expressed a truncated 53BP1 protein (53BP1-BRCTΔ, residues 1–1,711) that supports WT CSR frequencies38. As expected, WT 53BP1 expression efficiently rescued class switching in 53BP1−/− B cells, while the W1495A mutant could not (Fig. 5h and Supplementary Fig. 4), consistent with its inability to interact with DSB sites1. In contrast, 53BP1 F1553R-reconstitution rescued CSR to levels comparable to that seen with WT 53BP1 (Fig. 5h), indicating that TIRR binding is not involved in CSR regulation. This observation is in line with previous reports that increased expression of 53BP1 or RNF168 enhances PARPi sensitivity, but not CSR33, and reinforces the idea that 53BP1 stimulates productive CSR and inactivates mutagenic DNA repair via distinct pathways39. Together, our results with the F1553R mutant show that the impact of TIRR on 53BP1 activity is complex and context-dependent, but overall the TIRR–53BP1 interaction substantially influences genomic stability.
A nucleotide- and RNA-binding and processing enzyme can be engineered to bind 53BP1
While TIRR forms a tight complex with 53BP1 and regulates 53BP1 function in cells, it is not known how TIRR is removed from 53BP1 in response to DNA damage. Because TIRR is an RNA-binding protein and because it has been shown that small noncoding RNAs play a role in the recruitment of 53BP1 to DNA damage sites18,19,20,40, it is conceivable that RNA molecules may help displace 53BP1 from TIRR. Since the RNA targets of TIRR have not been identified, whereas the nucleic acid binding and RNA processing activities of NUDT16 are well understood, we engineered NUDT16 into a 53BP1-binding protein as a proxy to examine the possibility that nucleic acids interfere with 53BP1 binding. The TIRR 53BP1-binding loop was introduced in NUDT16, and NUDT16 Arg5 (corresponding to Lys10 in TIRR) was replaced by a lysine. Strikingly, isothermal titration calorimetry revealed that while NUDT16 has no affinity for 53BP1, the engineered protein (NUDT16 Tudor-interacting or NUDT16TI) binds 53BP1 with a Kd of 1.2 μM at 22 °C in the presence of 150 mM NaCl, which is similar to the affinity of TIRR for 53BP1 (Fig. 6).
Designer protein NUDT16TI recapitulates the 53BP1 binding mode of TIRR
The X-ray structure of NUDT16TI–53BP1, which was determined to 2.49-Å resolution (Fig. 6a, Table 1, and Supplementary Fig. 1b), reveals that we recreated a binding interface comparable to that in TIRR–53BP1, even though NUDT16 is substantially smaller than TIRR (Fig. 6a and Supplementary Fig. 5a,b). NUDT16TI is a homodimer, and in the crystal each protomer is bound to one 53BP1-Tudor molecule. Akin to the TIRR–53BP1 complex (Fig. 1b), the N termini of the two 53BP1-Tudor molecules both point toward the upstream oligomerization region of 53BP1 (Fig. 6a). In solution, a 400-residue 53BP1 homodimeric construct interacts with one NUDT16TI homodimer, as shown by sedimentation-velocity analytical ultracentrifugation (Fig. 6b). NUDT16TI is also a highly stable protein, allowing quantitative characterization of several mutants using isothermal titration calorimetry, which was not possible with TIRR due to its propensity to precipitate (Fig. 6c). In agreement with the NUDT16TI–53BP1 structure, the NUDT16TI R106S (R107 in TIRR) mutation abolished binding, while the K5S (K10 in TIRR) mutation, or deletion of His105 (His106 in TIRR), decreased the affinity by ~14- or ~38-fold, respectively (Fig. 6c). Selected mutations in 53BP1 (W1495A, D1521A, F1553R) also abolished 53BP1 interaction with NUDT16TI, further validating the binding interface identified in the crystal structure (data not shown).
Nucleotides and RNA molecules displace 53BP1 from designer protein NUDT16TI and TIRR
We showed that NUDT16TI retains WT affinity for its nucleic acid substrate IMP27 (Kd ~15 μM; Fig. 7a). It is noteworthy that the nucleotide-binding site in NUDT16 and NUDT16TI maps to a small area of a large, positively charged surface of TIRR that forms a plausible RNA-binding channel (Fig. 7a,b). Notably, an unbiased high-resolution mapping of RNA-binding regions in the nuclear proteome identified this channel and the adjacent 53BP1-binding loop (present study) as the RNA binding site of TIRR17, suggesting that 53BP1 and RNA could compete for the same binding region of TIRR (Fig. 7c).
Next, we asked whether IMP interferes with the NUDT16TI–53BP1 interaction. In the presence of excess IMP, the Kd of NUDT16TI for 53BP1 increased by roughly three-fold (Fig. 7a). This decreased affinity is readily explained by a steric effect, as the hairpin motif formed by residues 1,549–1,552 of 53BP1 points toward the IMP (and RNA) binding site within NUDT16TI (note Glu1551 in Fig. 7a). That a single nucleotide can substantially decrease the affinity of NUDT16TI for 53BP1 suggested that a longer RNA molecule could prevent binding of 53BP1 to TIRR, and indeed, RNA preparations from cell extracts blocked the interaction of 53BP1 with TIRR (Fig. 7d). In our current assay, RNAs extracted from irradiated or nonirradiated cells had indistinguishable effects. In the future, more elaborate investigations will need to be devised to possibly isolate ionizing radiation-specific RNAs that dissociate the TIRR–53BP1 complex. Our studies of TIRR and NUDT16TI provide a proof of principle for the involvement of RNAs in the recruitment of 53BP1 to DNA damage sites.
Our studies reveal in near-atomic detail a hitherto unobserved regulatory mechanism in which the PTM reader function of a protein, 53BP1, is inhibited by another protein, TIRR. We engineered a nucleotide- and RNA-binding and processing enzyme, NUDT16, to interact with 53BP1 (NUDT16TI) while preserving its nucleic acid binding properties. With this designer system, we unambiguously validated the TIRR–53BP1 binding mechanism by recreating this mechanism from first principles. The structures, validated by mutagenesis, guided the design of a separation-of-function 53BP1 mutant (F1553R) that cannot bind TIRR but retains affinity for chromatin. The hyperactive nature of 53BP1 F1553R highlights the important role of TIRR in maintaining 53BP1 in an inactive state in the absence of DNA damage.
The regulation of 53BP1 recruitment to DNA damage sites is complex, as it involves constitutive (H4K20me2) as well as DNA damage-induced (H2AK15ub) PTMs; factors like JMJD2A, L3MBTL1, RAD18, and RNF169 that compete with 53BP1 for reading these PTMs21,41,42,43,44,45,46; acetylation of H2AK15 that blocks 53BP1 recruitment47; and the direct inhibition of the 53BP1 PTM-reader function by TIRR14. As 53BP1 forms a tight complex with TIRR, determining how 53BP1 dissociates from TIRR is of utmost importance to understanding how 53BP1 is recruited to DNA damage sites. There is evidence that RAP1-interacting factor 1 (RIF1), which binds phosphorylated 53BP1 in response to DNA damage, is involved in the dissociation process14, but how this happens is not known. RIF1 has not been shown to directly bind TIRR. It was recently reported that noncoding RNAs control the DDR and contribute to 53BP1 recruitment18,19,20. 53BP1-Tudor has affinity for RNA molecules20,40, but it is unlikely that such RNAs would be involved in dissociating 53BP1 from TIRR, as the RNAs would be expected to mask the histone binding surface of 53BP1 required for its recruitment to damaged chromatin. As TIRR is both an inhibitor of 53BP1 and an RNA-binding protein, relief of 53BP1 inhibition by a TIRR-binding RNA molecule would be a simple and plausible mechanism. In addition, we note that RIF1 also binds RNAs17. In response to DNA damage, RIF1 interacts with phosphorylated 53BP1 in the soluble nuclear fraction and could possibly shuttle RNAs to TIRR, leading to its dissociation from 53BP1. Recombinant TIRR co-purifies with nucleic acids, and it requires nucleic acid removal by treatment with benzonase and purification at high salt concentration to bind 53BP1 (data not shown). The TIRR–53BP1 and NUDT16TI–53BP1 structures that we determined reveal overlap between their nucleic acid- and 53BP1-binding surfaces, further supporting the relief-of-inhibition mechanism. We showed that a single nucleotide diminishes NUDT16TI–53BP1 interaction and that purified cellular RNA displaces 53BP1 from TIRR.
The RNA sequence(s) recognized by TIRR are currently unknown, and its RNA-binding mechanism remains to be established. Notably, human cleavage factor Im (CFIm), a component of the pre-mRNA 3′ processing complex, contains a Nudix-like subunit (CFIm25) that, like TIRR, binds but does not hydrolyze RNA48. Nevertheless, the mechanism for TIRR RNA-binding cannot be inferred from the structure of the CFIm25–RNA complex48, as the two proteins do not share the same RNA-binding surfaces. Since the nuclear environment is rich in RNA, we speculate that TIRR recognizes a specific RNA motif with high affinity, which allows dissociation of the tight TIRR–53BP1 complex in response to DNA damage.
In summary, our work with TIRR and a designer protein provides proof of principle that, in response to DNA damage, an RNA molecule could trigger dissociation of TIRR from 53BP1, thereby promoting 53BP1 recruitment to DSBs. Work in progress is directed at further elucidation of this mechanism.
Plasmids for protein expression in E. coli
Human 53BP1 encompassing the tandem Tudor domain and the UDR sequence (residues 1,484–1,635) was cloned in a GST-based vector as reported21. A shorter 53BP1 construct containing only the Tudor domains (residues 1,484–1,603) was cloned in a modified pET vector encoding an N-terminal His6 tag cleavable by the tobacco etch virus (TEV) protease, as reported1. Human TIRR (residues 6–211), codon-optimized for E. coli expression, was cloned in a pBB75 vector with a noncleavable C-terminal His6 tag and without a tag. Another version of TIRR, not codon-optimized for E. coli expression, was cloned in a pET-based vector that adds a noncleavable C-terminal His6 tag to the protein. NUDT16 and NUDT16TI were cloned in pET-based vectors encoding an N-terminal His6 tag cleavable by TEV protease or PreScission protease or encoding a noncleavable C-terminal His6 tag. NUDT16TI was also cloned in pBB75 without a tag. All mutations were introduced by standard site-directed mutagenesis.
The proteins were expressed or co-expressed in BL21(DE3) E. coli cells grown at 37 °C in LB media to an OD600 of ~0.6 and induced with 0.5 mM isopropyl-β-d-thiogalactoside at 15 °C for ~16 h. Harvested cells were lysed using a microfluidizer (Avestin Emulsiflex C5), incubated with 250 U of benzonase (EMD Millipore) per 1 L of cell culture in the presence of 1.5 mM MgCl2 for 1 h at 4 °C, and then centrifuged. Supernatants were loaded into appropriate purification columns.
For the binding assays, His6-tagged 53BP1 was first purified by Ni2+-nitrilotriacetic acid (NTA) agarose chelation chromatography (Qiagen) using a standard solution of 50 mM sodium phosphate (NaPi), pH 7.5, and 300 mM NaCl with different imidazole concentrations to bind (5 mM), wash (20 mM), and elute (250 mM) the protein. The various versions of His6-tagged TIRR, NUDT16, and NUDT16TI were purified similarly, but at high salt, with 1 M and 500 mM NaCl during wash and elution, respectively. N-terminal His6 tags were next cleaved from the proteins by 4 °C overnight incubation with TEV protease or PreScission protease and 10 mM DTT. Samples were then concentrated and purified by size-exclusion chromatography using a HiLoad 16/60 Superdex 75 column (GE Healthcare) and a running buffer of 50 mM NaPi, pH 7.5, 500 mM NaCl. Samples were concentrated and buffer exchanged in 20 mM Tris-HCl, pH 7.5, with 300 or 500 mM NaCl.
To prepare the TIRR–53BP1 and NUDT16TI–53BP1 complexes, His6-tagged 53BP1 and untagged TIRR or NUDT16TI were co-expressed and co-purified by Ni+2-NTA (Qiagen) at high salt as above. After the tag on 53BP1 was cleaved (as above), the samples were purified by size-exclusion chromatography using the HiLoad 16/60 Superdex 75 column (GE Healthcare) and running buffer of 50 mM NaPi, pH 7.5, 500 mM NaCl. Samples were next incubated with ~100 μL Ni+2-NTA (Qiagen) for 30 min at 4 °C in a nutator, re-injected into the HiLoad 16/60 Superdex 75 column (GE Healthcare) with a running buffer of 25 mM HEPES, pH 7.5, 500 mM NaCl, and then concentrated in the presence of 2 mM TCEP.
For GST pull-down assays, GST and GST-tagged 53BP1 were first purified by passing through a GSTPrep FF 16/10 column (GE Healthcare), washing with PBS (pH 7.3), and eluting with PBS (pH 8.0) and 20 mM glutathione. On the other hand, GST-tagged 53BP1 co-expressed with C-terminally His6-tagged TIRR was first Ni+2-NTA (Qiagen)-purified at high salt, as above. Subsequently, all proteins were further purified by size-exclusion chromatography using HiLoad 16/60 Superdex 75 or 200 columns (GE Healthcare) and a running buffer of 50 mM NaPi, pH 7.5, 500 mM NaCl.
The nucleosome core particle ubiquitylated at histone H2A Lys15 and bearing a dimethylation mimic at histone H4 Lys20 (H4KC20me2; NCP-ubme) was prepared as previously reported21.
To probe the displacement of TIRR from 53BP1 by RNA molecules, 1 μg of purified recombinant Flag-tagged 53BP1 (residues 1,220–1,631) was bound to anti-Flag agarose beads (Sigma) and incubated with recombinant His6-tagged TIRR (1 μg) in TGEN150 buffer (20 mM Tris-HCl, pH 7.65, 150 mM NaCl, 3 mM MgCl2, 10% glycerol, and 0.01% NP-40) for 1 h at 22 °C on a roller. Identical aliquots of the washed slurry were next incubated in TGEN150 buffer with 200 ng and 1 μg of nuclear RNAs purified from untreated and from 5-Gy-irradiated HeLa cells, respectively, and without RNA as a control, for 1 h at 22 °C on a roller. The final volume of the incubation mixtures was 200 μL. The beads were washed five times with TGEN150 buffer and proteins retained on the beads were resolved and identified using 12% SDS/PAGE and western blot.
GST pull-down assay
GST pull-down assays were performed by first mixing 40 μL of 50% GSH slurry (Clontech) in buffer 1 (50 mM Tris-HCl, pH 8.0, 150 mM NaCl, 0.05% NP-40, 0.1% BSA) and bait (3 μg GST or equimolar amounts of mutant T1609E/S1618E GST-53BP1, and WT GST-53BP1, free or in complex with TIRR with C-terminal His6-tag) for 1 h on a nutator at 4 °C. Beads were then washed three times with buffer 1 (1 mL, 5 min), centrifuging (21,000 g, 2 min) between washes. Input NCP-ubme (36 μg) was added to the GST beads with immobilized baits and mixed for 2 h on a nutator at 4 °C. Beads were washed 5 × with buffer 1, removing NP-40 and BSA in buffer 1 in the last wash, and resuspended with 40 μL of 2 × Laemmli dye. Beads were boiled for 2 min and 10 μL of the supernatant was loaded onto a 4–20% TGX gel (Biorad). Protein bands were transferred onto a nitrocellulose membrane (Trans-Blot Turbo System, Biorad) and processed for western blot analysis. The membrane was blocked (5% nonfat milk in TBS, 1 h at room temperature), incubated with primary antibody (1:1,000 dilution in 1% BSA in TBS, overnight 4 °C), washed 5 × 5 min with TBST, incubated with HRP-conjugated secondary antibody (1:10,000 dilution in 1% nonfat milk in TBS, 1 h at room temperature), washed 5 × 5 min with TBST, and developed with an ECL reagent for imaging using a ChemiDoc MP system (Biorad).
The sources and dilutions for the antibodies used for western blot analysis of the GST pull-down assays are as follow: anti-ubiquitin (Cell Signaling P4D1, 1:1,000); anti-H2A (Millipore 07146, 1:1,000 dilution); anti-K15 ubiquitylated H2A (a gift from Z. Zhang, Mayo Clinic, 1:500 dilution); anti-GST (Santa Cruz sc-138, 1:1,000 dilution), anti-mouse HRP-conjugated (Cell Signaling 7076, 1:10,000 dilution), and anti-rabbit HRP-conjugated (BioRad 172-1019, 1:10,000 dilution).
Crystals were grown by the hanging-drop method, mixing 1.5 μL of the protein sample and 1.5 μL of the reservoir solution for the drop, and suspending 2–4 of these drops over 0.5 mL of reservoir solution. Crystals of apo-TIRR (residues 6–211) and TIRR bound to 53BP1 tandem Tudor domain (residues 1,484–1,603) were obtained using a sample of 10–20 mg/mL of co-expressed TIRR–53BP1 tandem Tudor domain and reservoir solution 1 (0.06 M citric acid, 0.04 M Bis-Tris propane, pH 4.1, 16% PEG 3,350) and solution 2 (0.1 M MES, pH 5.5, 0.2 M calcium acetate, 7% isopropanol), respectively. Crystals of NUDT16TI–53BP1 tandem Tudor complex were obtained using 15–20 mg/mL of co-expressed proteins and reservoir solution 3 (0.1 M sodium citrate tribasic dehydrate, pH 5.6, 0.2 M ammonium acetate, 10% PEG 4,000). All crystals were obtained at 22 °C.
Crystals of apo-TIRR, TIRR–53BP1 and NUDT16TI–53BP1 were cryoprotected in 25% PEG 3,350 in H2O, 30% glycerol in H2O, and 25% xylitol in solution 3, respectively, and were quick-frozen in CryoLoops (Hampton Research) with liquid nitrogen. The space group of the apo-TIRR crystals is P1, with two molecules per asymmetric unit. The space group of the TIRR–53BP1 crystals is P1 with two molecules of 53BP1 and four molecules of TIRR in the asymmetric unit. The space group of the NUDT16TI–53BP1 crystals is P43 21 2, with two molecules of 53BP1 and two molecules of NUDT16TI in the asymmetric unit.
X-ray diffraction data for apo-TIRR were obtained with a Rigaku Micromax-007/R-Axis IV2+ X-ray diffractometer. For TIRR–53BP1 and NUDT16TI–53BP1, data were collected at the 19-ID beamline of the Advanced Photon Source at Argonne National Laboratory, IL. All diffraction patterns were indexed, integrated, and scaled with HKL200051. To compensate for radiation damage sensitivity, data from four different crystals of TIRR–53BP1 were scaled and merged. Initial phases for apo-TIRR, TIRR–53BP1 and NUDT16TI–53BP1 were obtained by molecular replacement using the atomic coordinates of NUDT16L1 (PDB 3KVH), NUDT16 (PDB 2XSQ), and 53BP1 (PDB 2G3R) as search models in PHENIX52. The initial models were completed with manual building in COOT53 and refined in PHENIX. Statistics of the final structures are shown in Table 1. Molecular representations were generated with PyMol54.
Analytical ultracentrifugation (AUC)
Sedimentation-velocity measurements were performed at 30,000 or 40,000 r.p.m. using a Beckman Coulter Optima AUC instrument and an An-50 Ti rotor. Data were obtained after 15 h of centrifugation at 20 °C by monitoring the absorbance between sample and blank at 230, 280, or 290 nm. Protein samples were in 50 mM sodium phosphate buffer, pH 7.5, with 300 mM NaCl (Fig. 1c) or 100 mM NaCl (Fig. 6b). Data were analyzed using Sedfit55.
Isothermal titration calorimetry (ITC)
ITC experiments were carried out at 22 °C using an iTC200 calorimeter (MicroCal, Malvern). Samples were buffer-exchanged extensively with 20 mM Tris-HCl, pH 7.5, 20 mM MgCl2, 150 mM NaCl. Through the calorimeter injection syringe, 53BP1 at a concentration of 0.4 mM was delivered as a series of 2 μL injections every 3 min (iTC200) to the reaction cell containing NUDT16 or NUDT16TI (WT and mutants), at a concentration of 20 μM. To probe, the NUDT16TI–53BP1 interaction in the presence of inosine 5′-monophospate (IMP, Sigma), IMP was added in both the cell and injection syringe at 16 mM final concentration. For the titration of NUDT16 and NUDT16TI with IMP, IMP was at a concentration of 1.5 mM and NUDT16 or NUDT16TI was at a concentration of 50 μM. The measurements were paired with control titrations for heat of dilution. Data were analyzed using a one-site model with Levenberg–Marquardt nonlinear regression programmed in Origin 7.0 software (OriginLab Corporation).
Cell culture and antibodies
Cells were grown in Dulbecco’s modified Eagle medium (DMEM) containing 10% FCS. Parental cells were confirmed to be free of any mycoplasma contamination. Mouse antibodies were employed against Flag M2 (Sigma F1804), β-tubulin (Sigma T8328), cyclin A (Santa Cruz sc-271682), and γH2AX (Millipore 05-636). Rabbit antibodies were against TopBP1 (Bethyl Laboratories A300-111A), PTIP (Bethyl Laboratories A300-370A), 53BP1 (Cell Signaling 4937S), 53BP1 (Santa Cruz sc-22760), TIRR (Sigma HPA-044186), and phospho53BP1 (Ser25/29; Cell Signaling 2674S).
Cells were grown on glass coverslips, fixed in 4% formaldehyde in PBS for 15 min at room temperature, and blocked and permeabilized for 1 h in PBS containing 0.3% Triton X-100, 1% BSA, 10% fetal bovine (or 3% goat serum). Incubation with primary and secondary antibodies (Alexa Fluor, Molecular Probes) was conducted in PBS containing 1% BSA and 0.1% Triton X-100 for 1 h at room temperature. Coverslips were mounted using DAPI Fluoromount-G (SouthernBiotech).
FRAP analysis was performed on stably expressing GFP-53BP1-BRCTΔ WT and GFP-BRCTΔ F1553R HeLa cell lines. Cells were plated on 35-mm round glass-bottomed dishes 24 h before FRAP. Dishes were placed in an environmentally controlled closed-chamber system (Tokai Hit) that was maintained at 37 °C and 5% CO2 during imaging. The chamber system was mounted on an inverted confocal microscope (FluoView 1000, Olympus). A defined circular area in a single nucleus, termed region of interest (ROI), was photobleached using a 488-nm laser set at 80% power. Time-lapse images were acquired at an interval of 5 s at 10% laser power for 400 s. Average fluorescent intensities in the bleached region were normalized against intensities in the unbleached area of the same nucleus. The adjusted fluorescence intensity at each time point is represented as the fraction of the prebleach intensity at the ROI. The fluorescence intensity curve was plotted using the mean of 5 ROIs, each in a different nucleus in different fields.
Double immunoaffinity purification
TIRR harboring C-terminal Flag- and HA-epitope tags (TIRR-FH) was stably expressed in cells by retroviral transduction. Viral vector transduction, cell fractionation, and purification of the protein complexes were carried out as reported previously56. Proteins were identified using mass spectrometry conducted at Harvard Medical School Taplin Biological Mass Spectrometry Facility.
Cells were lysed for 30 min in 20 mM Tris-HCl, pH 7.65, 250 mM NaCl, 0.5% NP-40, 5 mM EDTA, 5% glycerol, and protease and phosphatase inhibitor cocktails (Roche). Protein concentrations from cleared supernatant were estimated using the Bradford dye-binding method (Biorad). We then incubated 500 μg of whole-cell extracts on a roller for 16 h at 4 °C with an anti-Flag antibody (Sigma). Resins were washed five times with TGEN150 buffer before elution in 0.1 M glycine, pH 2.9. Eluted proteins were analyzed by immunoblotting.
Cell viability assay
To measure the sensitivity of cells expressing 53BP1 F1553R to PARP inhibition by olaparib, 53BP1−/− MEFs (kindly provided by P. Jeggo, University of Sussex) were transduced with the retroviral pOZ vector, empty or containing the cDNA of 53BP1, with or without the F1553R mutation, and lacking the BRCT domains. Selected clones were transfected twice at 24-h intervals using Lipofectamine (Invitrogen) with a control siRNA (AAGCCGGUAUGCCGGUUAAGU) or an siRNA directed against BRCA1 (CAGCAGUUUAUUGCUCAUUGA). The cells were seeded into 96-well plates 24 h after the second transfection. The day after plating, olaparib (ChemieTek) was serially diluted in media and added to the wells. Five days later, olaparib was removed and cells were incubated in drug-free media for 72 h. The number of viable cells in culture was determined from the ATP-based CellTiter-Glo luminescent cell viability assay (Promega) using a luminescence microplate reader (CLARIOstar, BMG Labtech). For each olaparib concentration, data were plotted as a percentage of cell survival in drug-free media.
RPE1 (53BP1Δ) cells stably expressing 53BP1 (WT or F1553R mutant) were transduced with pWZL-TRF2ΔBΔM retrovirus (Addgene #18013)57. Seven days later, cells at 80% confluency were treated with 0.1 μg/mL colcemid for 2 h before trypsinization and neutralization in media. Cell pellets were incubated in 0.075 M fresh KCl for 10 min, then fixed in 3:1 methanol:acetic acid solution followed by one methanol:acetic acid wash. Metaphases were dropped on slides on a humidified heat block at 42 °C and air-dried. They were fixed in 4% formaldehyde, treated with 1 mg/mL pepsin for 10 min at 37 °C, fixed again in 4% formaldehyde, and dehydrated with sequential rinses in 70, 90, and 100% ethanol for 5 min each. Telomere PNA-FISH was performed with 20 nM Cy3-TelC probe (PNA Bio Inc.) in hybridization solution (0.5% blocking reagent (Roche), 70% formamide in 10 mM Tris-HCl, pH 7.2) overnight at 4 °C following 5 min denaturation at 80 °C. Following washes, slides were dehydrated in sequential ethanol wash and mounted in ProLong Gold. Images were captured on a Zeiss Imager Z1, and the percentage of chromosome fusions per chromosome was measured.
All experiments involved 8- to 10-week-old, sex-matched littermate control animals on an inbred C57BL/6 background. 53BP1−/− mice (MGI: 2654201) were generated and described elsewhere58. All experiments were approved by the University of Oxford Ethical Review Committee and performed under a UK Home Office License.
Primary B cell isolation, culture, and reconstitution
B cells were purified from red blood cell-lysed single-cell suspensions of four mouse spleens by magnetic negative selection using a B Cell Isolation Kit (Miltenyi Biotec). B cells (7.5 × 105 cells per well in a 6-well plate) were cultured in RPMI supplemented with 10% FCS, 100 U/mL penicillin, 100 ng/mL streptomycin, 2 mM l-glutamine, 1 × MEM nonessential amino acids, 1 mM sodium pyruvate, and 50 μM β-mercaptoethanol. B cells were stimulated with 10 ng/mL mouse recombinant IL-4 (Peprotech) and agonist anti-CD40 antibody (0.5 μg/mL; Miltenyi Biotec; FGK45.5). Cultures were grown at 37 °C with 5% CO2 under ambient oxygen conditions. Filtered retroviral supernatants, harvested 48 h after co-transfection of BOSC23 cells with 7 μg pCL-Eco and 7 μg pMX-DEST-GFP-derived plasmids were used to infect IL-4/anti-CD40-stimulated B cell cultures in the presence of polybrene (2.5 μg/mL) and HEPES (20 mM) by spinoculation (850 g for 90 min at 30 °C). After a rest period of 4–6 h, viral supernatants were removed and replaced with IL-4/anti-CD40-supplemented culture medium. Three days later, transduced B cells were analyzed using a FACSCanto (BD Biosciences); analysis was performed using FlowJo Software v10 (TreeStar). Cells were resuspended in PBS with 2% BSA and 0.025% sodium azide, blocked with Mouse BD Fc Block (1:500, BD Pharmingen 553141), and immunostained with biotinylated anti-mouse IgG1 (1:100, BD Pharmingen 553441; clone A85-1) and streptavidin APC (1:500, Thermo Fisher 17-4317-82). Live/dead cells were discriminated after staining with Zombie Aqua viability dye (1:200; BioLegend 423102). Surface IgG1 expression was determined in gated cell populations positive for the expression of an eGFP retroviral reporter.
Further information on experimental design is available in the Nature Research Reporting Summary linked to this article.
The statistical analyses were performed with GraphPad Prism 5.0 using a two-tailed Student’s t tests (Mann–Whitney test).
Publisher’s note: Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
We are very grateful to R. Alkire, N. Duke, and J. Lazarz at Argonne National Laboratory for their outstanding assistance. X-ray diffraction data were collected at Argonne National Laboratory, Structural Biology Center (SBC) at the Advanced Photon Source. SBC is operated by UChicago Argonne, LLC, for the US Department of Energy, Office of Biological and Environmental Research, under contract DE-AC02-06CH11357. This research was supported by NIH grants R01 CA132878, R01 GM116829, and P50 CA136393 (Mayo Clinic Ovarian Cancer SPORE developmental project) to G.M.; and by NIH grants R01 CA208244 and R01CA142698, DoD grant W81XWH-15-0564/OC140632, a Leukemia and Lymphoma Society Scholar grant, and the Claudia Adams Barr Program in Innovative Basic Cancer Research to D.C. M.V.B. was supported by DoD grant W81XWH-16-1-0391 and a Liz Tilberis award from the Ovarian Cancer Research Fund Alliance. G.C. received a Fellowship Award from the Mayo Clinic Cancer Center Fraternal Order of Eagles Funds. J.R.C. and C.O. were supported by a Cancer Research UK Career Development Fellowship Grant (C52690/A19270).