Multivalent binding of the hub protein LC8 at a newly discovered site in 53BP1

Abstract

Tumor suppressor p53 binding protein 1 (53BP1) is a scaffolding protein involved in poly-ADP ribose polymerase inhibitor hypersensitivity in BRCA1-negative cancers. 53BP1 plays a critical role in the DNA damage response and relies on its oligomerization to create foci that promote repair of DNA double-strand breaks. Previous work shows that mutation of either the oligomerization domain or the dynein light chain 8 (LC8)-binding sites of 53BP1 results in reduced accumulation of 53BP1 at double-strand breaks. Mutation of both abolishes focus formation almost completely. Here, we show that, contrary to current literature, 53BP1 contains three LC8-binding sites, all of which are conserved in mammals. Isothermal titration calorimetry measuring binding affinity of 53BP1 variants with LC8 shows that the third LC8-binding site has a high affinity and can bind LC8 in the absence of other sites. NMR titrations confirm that the third site binds LC8 even in variants that lack the other LC8-binding sites. The third site is the closest to the oligomerization domain of 53BP1, and its discovery would challenge our current understanding of the role of LC8 in 53BP1 function.

Significance

53BP1 is a key regulator of DNA damage repair, and this function leads to its involvement in a wide range of cancers. The DNA repair function of 53BP1 is dependent on the formation of nuclear bodies that decorate sites of double-strand breaks. Two LC8-binding sites in 53BP1 are important for ideal nuclear body formation. In this work, we identify a third LC8-binding site and investigate its interaction with LC8. The presence of a highly conserved third site underscores the need to revisit the role of LC8 in 53BP1 nuclear bodies.

Introduction

Faithful maintenance and replication of genetic information is critical for organismal fitness and viability. Given the importance of maintaining genomic integrity, cellular life has evolved complex mechanisms and signaling systems for protecting and repairing DNA from damage. Particularly insidious are double-strand breaks (DSBs), as these types of lesions are prone to initiate chromosomal rearrangements often resulting in the activation of oncogenes—or, conversely, the suppression of tumor-suppressor genes—leading to the expression of cancerous phenotypes (1,2). DSB repair is orchestrated by a complex signaling network known as the DSB response. A critical mediator of the DSB response is the scaffolding protein p53 binding protein 1 (53BP1), which was initially discovered via its interaction with the tumor suppressor p53 over 25 years ago (3). 53BP1 is involved in marking DSBs and protecting damaged DNA ends from end resection (2,3,4). 53BP1 plays a critical role in the DNA damage response, promoting repair of DSBs by forming phase-separated compartments at DSBs and recruiting DNA repair factors. Increased levels of 53BP1 are found in specimens from rectal carcinoma, skin cancer, and cervical cancer, and 53BP1 is associated with poor prognosis in breast and ovarian cancer (5). Importantly, loss of 53BP1 in individuals with BRCA-negative cancers results in loss of sensitivity to the chemotherapeutic poly-ADP ribose polymerase inhibitor (6).

53BP1 relies on a system of chromatin modification for recruitment to DSBs. Upon formation, DSB sites are flagged by ATM kinase via phosphorylation of histone variant H2AX, creating γH2AX. The adaptor protein, mediator of the DNA damage checkpoint, then binds to γH2AX and recruits ubiquitin ligases (7,8). This assembly recruits 53BP1 to chromatin, where it makes interactions with ubiquitinylated H2K15 and methylated H4K20 through its ubiquitin-dependent recruitment domain and tandem Tudor domains, respectively (9). 53BP1 forms large nuclear bodies at sites of DSBs, facilitated by a minimal focus-forming region (10), which includes the oligomerization domain (OD), Gly- and Arg-rich motifs, tandem Tudor domains, and ubiquitin-dependent recruitment domains (3). 53BP1 is also predicted to contain a long, disordered N-terminal tail that contains at least 28 sites for phosphorylation by ATM kinase and is thought to regulate recruitment of downstream effectors (2).

Close to the OD of 53BP1 is the dynein light chain 8 (LC8)-binding domain (LBD), which contains two known LC8-binding sites (11,12,13). LC8 is a 21 kDa dimeric protein that functions as a dimerization hub, binding and dimerizing over 100 client proteins at intrinsically disordered regions (IDRs) (14,15,16,17). LC8-binding motifs are eight residues long, including a three residue anchor sequence that is critical to interactions with client IDRs (Fig. 1 B) (18). The anchor sequence is most often composed of TQT residues, but the entire LC8-binding motif is quite variable (18,19). However, due to the sequence specificity of the anchor sequence, mutation of the three anchor residues to AAA results in complete loss of binding and does not alter the disordered state of the client protein (20,21). Because of the strong preference for a glutamine-threonine (QT) sequence in the LC8-binding site, LC8-binding sites are often referred to as QT motifs, or simply QTs. The LC8-binding sites in 53BP1 were originally identified by searching the sequence of 53BP1 for TQT and GIQ sequences in 53BP1, which were the only known consensus sequences for LC8 binding at the time (13). Since then, the structures of many LC8-client interactions have been solved (14,22,23,24), allowing for the development of an algorithm that predicts LC8 binding to IDRs based on the linear sequence of amino acids, volume, and polarity of residues in possible QTs (16,17). LC8Pred (19), a free online algorithm, uses this strategy to predict LC8 binding in disordered sequences. When using LC8Pred to analyze the sequence of human 53BP1 (Uniprot: Q12888), the two NMR-verified LC8-binding sites using short peptides (13) were correctly identified (1150–1157 and 1167–1174). However, LC8Pred also shows a high prediction score for a third site in 53BP1, located at 1192–1199.

Figure 1.

Domain architecture of 53BP1 and depiction of LC8 interaction. (A) Map of full-length 53BP1 domain architecture, along with domain map of the 53BP1 LC8-binding domain (LBD) sequence used in this study. LC8-binding sites (QT) are shown in blue, while QTs mutated to AAA to abolish binding are shown in red. (B) Structure of LC8 dimer (green) binding to two peptide partner. Dark blue residues on the peptide are the anchor of the binding site, while light blue residues are the rest of the QT (18) (PDB: 5E0L). (C) Model of multivalent LC8 interactions showing LC8 dimers (green) connecting each QT to the same QT on the adjacent client protein (46). (D) Schematic diagrams of 53BP1 LBDs in various species. Verified LC8-binding sites are shown in dark blue, predicted LC8-binding sites are shown in light blue, and the linkers are shown in green. Number of LC8-binding sites and linkers separating them are conserved in mammals, reptiles, some amphibians, and bony fish. Species names and sequence identifiers for each species shown in the figure are described in the materials and methods. (E) Sequences of LC8-binding sites identified by LC8Pred. The scores for the sequences and their locations are shown. Amino acid scores are generated from comparison of linear sequences of amino acids to known LC8-binding peptides, while volume and polarity scores from LC8Pred consider the properties of the amino acids in the sequence, with each position preferring amino acids with certain charge or bulkiness. No other sequences in 53BP1 besides those of QT1, QT2, and QT3 have favorable amino acid scores or volume and polarity scores. To see this figure in color, go online.

LC8 binds its clients in a parallel configuration, with the IDRs forming two beta strands at the binding interfaces of the symmetrical LC8 dimer (Fig. 1 B) (18). LC8 often interacts with its clients multivalently, resulting in a ladder-like assembly with LC8 dimers binding two client IDRs as in dynein intermediate chain, ASCIZ and Nup159 (Fig. 1 C) (20,22,25,26,27). The structural rigidity imposed on client peptides by LC8 binding can extend beyond the binding interface such that the entire assembly becomes rigid and rod-like. These complexes have almost always been presumed to be in register, in which a QT on a single client molecule is linked to the same QT on the other client molecule through a single LC8 dimer (Fig. 1, B and C). It is common for LBDs to be found near self-association domains, similar to the one seen in 53BP1 (28). These self-association domains are most commonly coiled coils and are usually dimeric as in the Swallow protein (29). In 53BP1, the OD is predicted to fold as a beta strand, which is unique for LC8 partners (30,31). Although structures of various 53BP1 fragments have been determined (32), no studies are available on 53BP1 LBD binding to LC8.

While LC8 is shown to reduce 53BP1 nuclear body formation and is thought to aid in structuring of the N-terminal tail of 53BP1 (12), how it does that is totally speculative. Here, we provide the first biophysical characterization of the 53BP1 LBD and its interaction with LC8. We demonstrate that 53BP1 binds three LC8 dimers instead of two and that, together, they form a rigid duplex preceding the OD. Since functional studies have been performed on constructs that eliminated only two of the three LC8 sites, our findings highlight the need to revisit the functional roles of 53BP1-LC8 interactions.

Materials and methods

Cloning, protein expression, and purification

Studies were carried out using human 53BP1 (Uniprot: Q12888) and human LC8-2 (Uniprot: Q96FJ2). A construct containing 53BP1 residues 1140–1225 in frame with an N-terminal hexahistidine tag and tobacco etch virus (TEV) protease cleavage site was used. Cysteine residues in the sequence were mutated to serine to avoid the formation of disulfide-induced aggregates. The sequence was codon optimized for expression in Escherichia coli in a pET24d expression vector (purchased from GenScript, Piscataway, NJ, USA). To abolish LC8 binding, the three anchor residues for each QT site were mutated to AAA amino acids using New England Biolabs’ (Ipswich, MA, USA) site-directed mutagenesis kit and custom primers. The wild-type 53BP1 LBD contains all three LC8-binding sites, while the QT variants contain only the indicated LC8-binding sites (i.e., QT1 contains intact LC8-binding site 1, while LC8-binding sites 2 and 3 are mutated to abolish binding). The wild-type LBD, QT1, QT2, QT3, QT1,2, and QT1,3 and an LC8-binding null mutant were used to characterize each site and to investigate cooperativity between sites.

All constructs were transformed into E. coli Rosetta DE3 cells and expressed in autoinducing media at 37°C for 24 h or Luria-Bertani for 4 h with 1 mM IPTG. For NMR experiments, cells were grown in either autoinducing minimal media with ¹⁵N or MJ9 minimal media supplemented with ¹³C-glucose and/or ¹⁵NH₄Cl. Cells were harvested and purified on Talon His-Tag Purification Resin (Takara Bio, Mountain View, CA, USA). The hexahistidine tag was removed from LC8 using TEV protease for NMR experiments. Since there is a 26-residue-long linker between QT1 (the most N-terminal QT) and the hexahistidine tag, the hexahistidine tag was not expected to alter interactions with LC8 and was not removed for other experiments. The tag was removed from 53BP1 for NMR experiments to reduce overlap in the spectra, and the sample was purified using size-exchange chromatography. Cleavage with TEV protease leaves an N-terminal cloning artifact with the residues Gly Ala His (GAH) preceding the desired sequence. The purity of recombinant proteins was over 95%, as analyzed by SDS-PAGE. Proteins were stored at 4°C and used within 1 week. Protein concentration was quantified using absorbance at 205 nm since the 53BP1 LBD sequence does not contain any residues that absorb strongly at 280 nm. Extinction coefficients at 205 nm for each construct are as follows: 53BP1 LBD with tag = 395,890 M⁻¹cm⁻¹, 53BP1 LBD without tag = 282,920 M⁻¹cm⁻¹, and LC8 without tag = 374,720 M⁻¹cm⁻¹ (33).

Circular dichroism

Spectra were recorded on a JASCO J-810 spectropolarimeter using a 1 mm cell. Protein samples were dialyzed overnight in 2 L of 20 mM sodium phosphate (pH 7.2) prior to data collection. Spectra were collected at 25°C at a protein concentration of 10 μM. Data shown are the average of 5 spectra, and results are reported in mean residue molar ellipticity (deg^∗cm²/dmol).

Isothermal titration calorimetry

Thermodynamics of the 53BP1 LBD-LC8 interaction were measured at 25°C using a VP-isothermal titration calorimetry (ITC) microcalorimeter (MicroCal, Northampton, MA, USA). The binding buffer contained 50 mM sodium phosphate and 150 mM sodium chloride (pH 7.2). In each experiment, LC8 at a concentration of 250–400 μM was titrated into an LBD construct. Cell concentrations were as follows: LBD, 10 μM; LBDQT1, 35 μM; LBDQT2, 35 μM; LBDQT3, 30 μM; LBDQT1,2, 15 μM; and LBDQT2,3, 15 μM. The LBD-binding null mutant was titrated as the other constructs, with 360 μM LC8 and 30 μM LBD-binding null mutant. Peak areas were integrated, and the data were fit to a single-site binding model in Origin 7.0. From this, stoichiometry (N), dissociation constant (K_D), change in enthalpy (ΔH), and entropy (ΔS) were determined. Reported data are the average of multiple runs. Error reported is the standard deviation of the data acquired.

Size-exclusion chromatography

Size-exclusion chromatography was performed on an S200 Superdex gel filtration column using an AKTA-FPLC (GE Healthcare, Chicago, IL, USA). Data for free 53BP1 LBDs were collected by injecting protein at 3 mg/mL in the ITC buffer described above. For each additional experiment, the 50 μM 53BP1 LBD was mixed with an appropriate amount of LC8 to achieve the molar equivalency desired. Samples were run at a flow rate of 0.6 mL/min. Expected masses are as follows: 53BP1 LBD, 12.3 kDa; LC8, 21.2 kDa; and 53BP1-LC8, 88.2 kDa. Proteins were detected using Optilab refractive index detector (Wyatt Technology, Santa Barbara, CA, USA).

Sedimentation velocity analytical ultracentrifugation

Analytical ultracentrifugation (AUC) experiments were performed using a Beckman Coulter Optima XL-A ultracentrifuge equipped with absorbance optics (Brea, CA, USA). For sedimentation velocity AUC (SV-AUC) experiments, samples were loaded into epon2-channel sectored cells with a 12 mm optical pathlength and run at 42,000 Rpm in a four-cell Beckman Coulter AN 60-Ti rotor at 20°C. Scans were performed continuously for a total of 300 scans per cell. Data were fit to a continuous c(S) distribution using the software SEDfit (34). Sedimentation coefficients are expressed in Svedbergs (S).

NMR

NMR experiments were performed on a Bruker 800 MHz Avance III HD spectrometer (Bruker Biosciences, Billerica, MA, USA) equipped with a 5 mm TCI cryoprobe with Z-axis gradient (Bruker). NMR experiments were carried out at 10°C in 20 mM sodium phosphate, 50 mM sodium chloride, and 1 mM Sodium azide (pH 6.5), with 10% D₂O, a protease inhibitor mixture (Roche Applied Science, Madison, WI, USA), and 2-2 dimethylsilapentane-5-sulfonic acid for chemical shift referencing. For backbone assignments, HNCACB, HN(CO)CACB, HNCO, and HN(CA)CO experiments were acquired with nonuniform sampling on the ¹³C and ¹⁵N uniformly labeled 53BP1 LBD. NMR data were processed using nmrPipe (35) and non-uniform sampling artifacts removed using SMILE (36). Assignments were obtained using NMRViewJ (37). Steady-state ¹H-¹⁵N heteronuclear nuclear Overhauser effects ({¹H}-¹⁵N NOE) experiments were collected with an 8 s saturation time. Error bars were calculated using σ/(NOE) = [(σI_sat/I_sat)² + (σI_unsat/I_unsat)²]^1/2, where I_sat and σI_sat are the intensity of the peak and its baseline noise. R1 and R2 data were collected with heteronuclear single quantum coherence (HSQC)-based temperature compensated pulse sequences (38). Time points were collected in triplicate for error estimation. To analyze the interaction of 53BP1 with LC8, ¹H-¹⁵N HSQC spectra were taken with 0.25, 0.5, 1, 2, 3, and 4 molar equivalents of unlabeled LC8 titrated into the ¹⁵N-labeled LBD using the NMR conditions described above. LBD, QT1,2, or QT3 samples were at a 50–100 μM concentration range, and LC8 at 750–2,000 μM was added to minimize dilution effects. Chemical shift indexing was performed using sequence-corrected shifts for amino acids in random coils (39,40,41).

Protein BLAST search and LC8Pred

Human 53BP1 (Uniprot: Q12888) was used to search protein BLAST for similar proteins (42). The sequence for representative species (Species, NCBI reference sequence; Rattus Norvegicus, NCBI: XP_006234927.1; Vombatus ursinus, NCBI: XP_027725777.1; Gallus Gallus, NCBI: XP_004944052.2; Chelonia mydas, NCBI: XP_043379238.1; Xenopus laevis, NCBI: XP_018108636.1; Rhinatrema bivittatum, NCBI: XP_029430825.1; Amia calva, NCBI: MBN3308935.1; Danio rerio, NCBI: NP_001073639.1; Styela clava; NCBI: XP_039265801.1; Rad9 in Saccharomyces cerevisiae, Uniprot: 14737) was downloaded in FASTA format and run on the LC8-binding prediction algorithm LC8Pred.

Results

The LBD of 53BP1 is fully disordered

We analyzed the sequence of human 53BP1 (Uniprot: Q12888) using LC8Pred (19), which accurately predicted the two previously described LC8-binding sites (anchored at residues 1155 and 1172) (13) and an additional third site anchored at residue 1197 (Fig. 1, A and E). This site has the highest LC8-binding score in the prediction and also lies between the verified binding sites and the OD. The entire sequence of the LBD is predicted to be disordered, a prerequisite for binding by LC8 (43). A BLAST search and LC8Pred prediction on the resulting sequences revealed that LC8 binding to 53BP1 is a conserved trait. Both the number of LC8-binding sites and approximate length of linkers separating them are conserved in mammals, birds, and reptiles (Fig. 1 D). To characterize the full LBD, we created a construct of human 53BP1 containing residues 1140–1225 just preceding the OD of residues 1231–1279.

Far-UV circular dichroism shows a spectrum with a strong negative ellipticity at 200 nm, which is consistent with a primarily disordered structure (Fig. 2 A). A ¹H-¹⁵N HSQC spectrum of the free 53BP1 LBD uniformly labeled with ¹⁵N also shows limited peak dispersion (Fig. 2 B). 80 out of 85 nonproline residues were assigned using data from 3D HNCO, HN(CA)CO, HN(CO)CACB, and HNCACB experiments. ΔCα-Cβ chemical shift indexing shows values between −0.5 and 0.5 for most residues (Fig. 2 C). The lack of contiguous regions with deviation from Cα-Cβ shifts from expected values indicates that the 53BP1 LBD is dynamic and disordered, consistent with the prediction. ¹⁵N relaxation experiments confirm that the 53BP1 LBD is disordered with little evidence of restricted motion across the chain (Fig. 2 C). ¹H-¹⁵N-NOE values of less than 0.5 for all residues (average = 0.26) indicate that the entire peptide is dynamic and unstructured.

Figure 2.

Structure of 53BP1 LBD. (A) Far-UV circular dichroism of 53BP1 LBD showing high negative ellipticity at 200 nm, characteristic of disordered proteins. (B) ¹H-¹⁵N HSQC spectrum of 53BP1 LBD taken at 800 MHz at 10°C in 20 mM sodium phosphate (pH 6.5) and 50 mM sodium chloride with 10% D₂O. Assignments for residues are labeled without the thousands place, which is 1--- for every residue. Add 1000 to each label for the residue number in full-length 53BP1. (C) Chemical shift indexing and dynamics measurements. Values of ΔCα-Cβ close to zero are indicative of lack of secondary structure, while contiguous positive values indicate alpha helices and contiguous negative regions indicate beta strands. The 53BP1 LBD shows no stretches of positive or negative chemical shift differences, indicating the absence of a stable secondary structure. R1, R2, and {¹H}-¹⁵N NOE data show evidence for primarily disordered protein, except for few higher values in R2 indicating motional restriction. Gaps are unassigned residues in the sequence. Error bars for {¹H}-¹⁵N NOE were calculated using σ/(NOE) = [(σI_sat/I_sat)² + (σI_unsat/I_unsat)²]^1/2, where I_sat and σI_sat are the intensity of the peak and its baseline noise. Time points in R1 and R2 measurements were collected in triplicate for error estimation. To see this figure in color, go online.

53BP1 binds LC8 cooperatively

To investigate the interaction between 53BP1 and LC8, we used size-exclusion chromatography (Fig. 3 A). Titration of 53BP1 with LC8 results in a peak that migrates earlier and reaches a maximum shift at 4 molar equivalents of LC8. A similar trend is seen by SV-AUC (Fig. 3 B). In SV-AUC, the free LBD has a sedimentation value of 1 S. Addition of LC8 results in a peak that shifts with increasing LC8 and reaches a maximum sedimentation of 4.9 S with 4 molar equivalents of LC8. The formation of a single peak with high molecular weight that migrates with high molar equivalents of LC8 suggests that binding to all sites is cooperative and does not populate a stable intermediate.

Figure 3.

Interaction of 53BP1 with LC8. (A) Titration of 53BP1 LBD with LC8 analyzed by size-exclusion chromatography. 53BP1 forms a single, monodisperse peak, and upon addition of LC8, a peak corresponding to a large mass migrates at lower volume. (B) A titration of 53BP1 LBD with LC8 analyzed by sedimentation velocity analytical ultracentrifugation (SV-AUC). 53BP1 forms a single peak with low sedimentation coefficient (S). Addition of LC8 results in a shift of the peak maximum to larger S value, and titration of LC8 results in peak migration to a maximum value of 4.85 S. To see this figure in color, go online.

ITC confirms a third LC8 binding site in 53BP1

ITC experiments on the 53BP1 LBD show that 53BP1 binds LC8 with submicromolar affinity (0.28 μM) and a stoichiometry of 3 (Fig. 4 A). The enthalpy and entropy of binding are both favorable, but the interaction is strongly enthalpically favored. We mutated all three QT anchor sites to AAA sequences to generate an LC8-binding null variant of the 53BP1 LBD. This peptide shows only weak signal by ITC when titrated with LC8 (Fig. 4 B). This is consistent with nonspecific binding of LC8 to a disordered sequence and indicates that the 53BP1 LBD binds LC8 specifically only at the three sites predicted by LC8Pred.

Table 1.

Thermodynamic parameters determined by isothermal titration calorimetry at 25°C

Sample	N	Overall KD (μM)	Overall ΔH (kcal/mol)	−TΔS (kcal/mol)	ΔG (kcal/mol)
LBD	3 ± 0.2	0.3 ± 0.2	−8 ± 1	−1 ± 1	−9 ± 0.3
QT1	1.3 ± 0.04	5.2 ± 0.9	−5.7 ± 0.2	−1.5 ± 0.3	−7.2 ± 0.1
QT2	1.2 ± 0.08	0.62 ± 0.06	−12 ± 0.3	3.5 ± 0.3	−8.5 ± 0.1
QT3	1.1 ± 0.04	1 ± 0.3	−4.9 ± 0.4	−3.4 ± 0.6	−8.2 ± 0.2
QT12	2.2 ± 0.007	0.18 ± 0.004	−9.2 ± 0.01	−0.08 ± 0.02	−9.3 ± 0.03
QT23	2.1 ± 0.01	0.19 ± 0.009	−8.4 ± 0.2	−0.8 ± 0.2	−9.2 ± 0.02
Null	N/A	N/A	N/A	N/A	N/A

Figure 4.

Thermodynamics of 53BP1-LC8 interactions by isothermal titration calorimetry. (A–G) Isothermal titration calorimetry of wild-type LBD (A), binding null mutant (B), QT1 (C), QT2 (D), QT3 (E), QT1,2 (F), and QT2,3 (G). A schematic of the LBD along with QT variants is shown on top of each isotherm. Blue QT indicates that the site was not mutated, and red QT indicates that that anchor was mutated to AAA to abolish binding for that site. Thermodynamics data were collected at 25°C in 50 mM sodium phosphate and 150 mM sodium chloride. (H) Bar graph of thermodynamic parameters measured in (A)–(G) and shown also in Table 1. To see this figure in color, go online.

To determine the effects of each binding site on the overall interaction, ITC was performed with single-site variants of the 53BP1 LBD (QT1, QT2, QT3) (Fig. 4, C–E). Each construct binds with a stoichiometry near 1, indicating that two chains of the LBD bind only to a single LC8 dimer. QT2 and QT3 bind LC8 with similar affinity (K_D is 0.65 μM for QT2 and 0.95 μM for QT3). However, QT1 binds with much lower affinity (K_D = 5.2 μM). This suggests that QT2 and QT3 are primarily responsible for the strong affinity of the 53BP1 LBD to LC8 and that QT1 may have some other role, such as tuning the affinity of the interaction through cooperative effects. Interestingly, the mechanism of each binding site varies dramatically. While each QT motif binds with strongly favorable enthalpy, the entropy of the reactions is very different. QT2 has the highest enthalpic contribution (−12 kcal/mol). It is also the only site with an entropic penalty (TΔS = 3.5 kcal/mol). QT3 has the lowest enthalpy (−4.9 kcal/mol) and the highest favorable entropy (TΔS = −3.4 kcal/mol), which compensates for this, resulting in a reaction with an affinity close to that of QT2. Comparatively, QT1 has modest enthalpy (−5.7 kcal/mol) and entropy (TΔS = −1.5 kcal/mol), resulting in the weakest overall interaction.

ITC measurements of double-site mutants are consistent with cooperative binding. Both QT1,2 and QT2,3 mutants have an affinity even higher than the wild-type protein (0.18 and 0.19 μM, respectively) (Fig. 4, F and G). QT1,2 has a higher enthalpy than QT2,3 (−9.2 and −8.4 kcal/mol, respectively), but a more favorable entropy in QT2,3 compensates for the comparatively low enthalpy.

NMR confirms the location of the three LC8-binding sites in 53BP1

To further investigate cooperativity in 53BP1-LC8 interactions, we conducted titrations of unlabeled LC8 into ¹⁵N-labeled 53BP1. Upon addition of LC8, global peak attenuation is observed (Fig. 5 A). This is largely due to the significantly increased size of the complex when bound to LC8 (88 kDa for the fully bound complex), leading to increased rotational correlation times and global peak broadening. Residues of 53BP1 that maintain their disorder display local tumbling, which reduces broadening such that these resonances are more likely to remain detectable. In the titration of the 53BP1 LBD, all residues close to LC8-binding sites lose intensity, indicating that the 53BP1-LC8 complex is either rigid compared with the free 53BP1 LBD or sampling multiple conformations in intermediate exchange, causing peak broadening at and near LC8 sites (Fig. 5 A). The uniform peak loss suggests cooperative binding since, even at low LC8 concentrations, we see loss of intensity of peaks near all the QT regions. The presence of three LC8 binding sites in 53BP1 is supported by the loss of resonances for residues between 1144 and 1211. In the fully bound 53BP1-LC8 complex, all resonances that are still detectable are found at both termini, suggesting that these residues maintain some degree of flexibility (Fig. 5 A).

Figure 5.

Identifying the LC8-binding sites on 53BP1 by NMR. Titration of uniformly labeled ¹⁵N 53BP1 LBD (A), QT1,2 (B), and QT3 (C) with LC8. Intensity ratios of the bound relative to the free are plotted as bar graphs as a function of residue number and also as points, corresponding to the average of peak intensities within the site, as a function of LC8 concentration. Peak disappearance is seen primarily in the regions corresponding to intact LC8-binding sites. Above each bar graph, the domain map of 53BP1 LBD is shown: functional QTs are in blue, while QTs mutated to AAA are in red. Average values for signal intensity ratio (I_bound/I_free) for each LC8-binding site are plotted above each domain map as a function of increasing LC8 concentration. Gaps in the sequence are unassigned residues. For the QT1,2 (B) and QT3 (C) titrations, there are gaps for the anchor motifs and some of the residues bordering the anchor motifs, which were not assigned for the mutants. Instead, for the QT variants, it was assumed that the peaks that do not match with the wild-type are from the mutated residues. These peaks were averaged and are shown above each titration. For all constructs, peak attenuation of AAA motifs is minimal, suggesting that these residues do not bind LC8. To see this figure in color, go online.

Titrations of QT1,2 and QT3 were conducted to confirm the regions of 53BP1 that bind to LC8. QT1,2 show peak attenuation of residues in and near sites 1 and 2 (Fig. 5 B). Mutation of the QT3 anchor (residues 1196–1198) to alanine results in chemical shift alterations for residues 1194–1201. To report on their change in intensity during titration of the QT1,2 variant, we considered the new peaks in the QT1,2 spectrum to be from the mutated residues and thus averaged their peak intensity at different titration points. These residues retain more signal intensity than the intact QT sites, indicating no binding in this region. Additionally, more residues in the C-terminus of the QT1,2 construct retain high signal intensity compared with the LBD, further supporting the absence of binding when the anchor residues are mutated to AAA. Overall, we see peak attenuation for 1148–1191, supporting that QT1 and QT2 both bind to the QT1,2 construct. Conversely, in the titration of QT3 (Fig. 5 C), significant peak attenuation is seen only for residues 1181–1204, which correspond to the QT3 region, indicating that LC8 binds 53BP1 specifically at QT3. It is worth noting that the signal loss extends significantly beyond the C-terminal end of QT3 and that the peak attenuation for QT1,2 and QT3 covers overlapping regions.

LC8 binding is conserved in mammalian 53BP1

Sequence analysis of LC8-binding sites in 53BP1 of various species shows interesting observations. 53BP1 has orthologs in yeast, Rad9 in S. cerevisiae, and Crb2 in S. pombe (2,3). While the sequence of these orthologs is not well conserved, several key features and their architecture, such as a disordered N-terminal tail, DNA-binding Tudor domain, and BRCT domains, are conserved (3,44). To determine whether LC8 binding is a conserved feature of 53BP1, we performed a protein BLAST search using human 53BP1 and LC8Pred (19,42) on the resulting sequences. In many species, LC8Pred predicts one to three LC8-binding sites within a short, disordered sequence of up to 52 residues. The 53BP1 LBD is highly conserved in mammals, with three LC8 binding sites and similar linker lengths separating them (Fig. 1 D). Interestingly, a slightly increased linker length is observed between QT2 and QT3 in marsupial mammals, which is conserved in avian and reptilian 53BP1 (Fig. 1 D). Caecilian 53BP1 shares architecture with reptilian 53BP1, but frog 53BP1 contains only two LC8-binding sites with linker lengths close to the mammalian linker between QT1 and QT2. In bony fish, 53BP1 has three LC8-binding sites, but the length of the linkers greatly diverges from mammalian 53BP1. Ray-finned fish and tunicate 53BP1 contain only a single LC8-binding site, which is similar to the 53BP1 ortholog in S. cerevisiae, which only has a single predicted LC8-binding site. All of this suggests that LC8 binding is an evolutionarily conserved feature of 53BP1. However, the number of LC8-binding sites in 53BP1, while conserved from reptiles to mammals, has increased since common ancestors of ray-finned fish. We also see a conservation in the linker length between LC8 binding sites in mammals, birds, reptiles, and some amphibians, suggesting that the length of these linkers is vital to proper regulation of 53BP1 dimerization.

Discussion

53BP1 is a key regulator of DNA damage repair, and its oligomerization is necessary for formation of 53BP1 nuclear bodies (10). In its long, disordered domain preceding the OD, there are three LC8-binding sites, including one first identified and described here. LC8 is known to alter the affinity of nearby dimerization domains (28), and is necessary for optimal 53BP1 nuclear body formation (12). Multivalency in LC8 binding in 53BP1 appears to be evolutionarily conserved through at least the reptiles, and there is at least one LC8-binding site as far as S. cerevisiae, suggesting that LC8 binding is an ancient and important part of 53BP1 function (Fig. 1 D). In all complexes of LC8 studied to date, LC8 either binds weak dimers, promoting their dimerization, or creates a scaffold onto which other partners can bind with higher affinity (20,25,29). LC8 can also restrict the conformational ensemble of dynamic complexes as we demonstrated in the RavP-LC8 complex (45). RavP contains an N-terminal dimerization domain that is separated from the C-terminal domain necessary for polymerase function by a long disordered linker. Since RavP contains a strong dimerization domain, we proposed that the function of LC8 is to restrict the conformation of the linker and align the C-terminal domains of RavP. Such a function for LC8 in 53BP1 interactions has been proposed already in which LC8 interactions are needed to align the N-terminal tails of 53BP1 to assist with recruitment of downstream DNA repair proteins (12). Our findings support this model and further identify a new site for LC8 binding that could broaden the role of LC8 in 53BP1.

First, we show by ITC that 53BP1 does contain three LC8-binding sites in contrast to what is known in the literature (11,12,13). In ITC experiments for each single QT, the binding is primarily driven by enthalpy, and the entropy of each reaction results in tuning of the binding affinity for each site. We show that the newly discovered LC8-binding site in 53BP1 binds LC8 in the absence of other binding sites and with submicromolar affinity (Fig. 3 C).

Additionally, the titrations of the 53BP1 LBD with LC8 analyzed by size-exclusion chromatography and AUC show that a large complex is formed early in the titration at low molar equivalents of LC8 (Fig. 3, A and B). This suggests that 53BP1 binds LC8 cooperatively such that once a single LC8 dimer binds an LBD pair, the subsequent binding events fill other LC8 binding sites with increased affinity. We also see uniform peak attenuation by NMR titration at all LC8-binding sites, further supporting cooperative binding in 53BP1-LC8 interactions.

The presence of a third QT in 53BP1 complicates previously published experiments (11,12) that mutated only two LC8-binding sites (QT1 and QT2). Given that LC8 can bind QT3 in the absence of other sites, and with minimally reduced affinity, it is likely that LC8 was still present in 53BP1 nuclear bodies in these experiments. Another experiment in which the full LBD was deleted from 53BP1 determined LC8 to be dispensable for nuclear body formation. However, in this study, a cry2-tag that oligomerizes when stimulated with light was used to initiate phase separation of 53BP1. This tag effectively reproduces the function of LC8 in the experiment, making it difficult to discount a critical role of LC8 in 53BP1 focus formation. Future work will elucidate the role of LC8 in 53BP1 nuclear body formation, and our identification of a third LC8-binding site underscores the expanding role of LC8 in 53BP1 in general.

Second, we show that LC8 interaction alters the dynamic behavior of the disordered 53BP1 LBD. In NMR titration experiments, the wild-type LBD shows peak attenuation upon binding to LC8 for residues 1144–1211. Mutation of QTs results in alterations of patterns of peak attenuation observed. The titration of QT1,2 shows peak attenuation only for residues 1148–1191, while the titration of QT3 shows peak attenuation for residues 1181–1204. The peak attenuation observed in these titrations is due to either increased rotational correlation time associated with rigidification of these regions of the LBD or intermediate exchange processes. It is well established that clients of LC8 form beta strands along the binding interface (21,23). In multivalent assemblies of LC8, the entire LC8-binding region of the client can become a rigid and rod-like structure (25), though this is not always the case (21).

Since the LBD of 53BP1 is close to its OD, the chains of the LBD may already be associated before interacting with LC8, as is the case with RavP, where a single LC8 binds to preassociated chains (45). Given the similarity with RavP in LC8 binding to preassociated chains, it is likely that these proteins share a similar function for LC8 binding, which is alignment of other domains in the protein to better facilitate the biological function of the client protein. We find that 53BP1 and LC8 interact with submicromolar affinity and that the interaction is cooperative such that no stable intermediates are formed at low LC8 concentrations. We speculate that the entire assembly of the LBD and LC8 becomes a rigid, rod-like structure that helps to regulate the binding of other 53BP1 binding partners to the N-terminal tail of 53BP1, such as Rif1. The proximity of QT3 to the OD may also extend the section of 53BP1 that becomes rigid, further regulating the binding of 53BP1 binding partners. Future work will be needed to determine the structure of the entire assembly and test this hypothesis.

Author contributions

Conceptualization, J.H. and E.B.; resources, P.R. and E.B.; data collection, J.H. and A.W.; visualization, J.H. and P.R.; formal analysis, J.H. and P.R.; methodology, J.H., P.R., and E.B.; writing – original draft, J.H.; writing – review & editing, J.H., A.W., P.R., and E.B.; funding acquisition, E.B.; supervision, E.B.

Editor: Alemayehu Gorfe.