Structure and Mechanism in Non-Homologous End Joining

Abstract

DNA double-stranded breaks (DSBs) are a particularly challenging form of DNA damage to repair because the damaged DNA must not only undergo the chemical reactions responsible for returning it to its original state, but, additionally, the two free ends can become physically separated in the nucleus and must be bridged prior to repair. In nonhomologous end joining (NHEJ), one of the major pathways of DSB repair, repair is carried out by a number of repair factors capable of binding to and directly joining DNA ends. It has been unclear how these processes are carried out at a molecular level, owing in part to the lack of structural evidence describing the coordination of the NHEJ factors with each other and a DNA substrate. Advances in cryo-Electron Microscopy (cryo-EM), allowing for the structural characterization of large protein complexes that would be intractable using other techniques, have led to the visualization several key steps of the NHEJ process, which support a model of sequential assembly of repair factors at the DSB, followed by end-bridging mediated by protein-protein complexes and transition to full synapsis. Here we examine the structural evidence for these models, devoting particular attention to recent work identifying a new NHEJ intermediate state and incorporating new NHEJ factors into the general mechanism. We also discuss the evolving understanding of end-bridging mechanisms in NHEJ and DNA-PKcs’s role in mediating DSB repair.

1. Introduction

DNA Double-stranded breaks (DSBs) are widely considered to be the form of DNA damage with the greatest potential for severe cytotoxicity [1, 2]. Defective DSB repair can lead to profound aberrations of genetic material that can result in either apoptosis or carcinogenesis [3]. Unlike other forms of DNA damage in which the lesion is generally limited to one strand of DNA, allowing for repair to proceed directly following recognition of the damaged DNA, DSBs pose an additional problem in that the two ends of DNA must be brought back into proximity to one another prior to repair, a nontrivial task in the crowded nuclear environment. The myriad sources of DSBs further complicate their repair by producing a variety of different damaged termini that cannot be directly ligated [4, 5] (Fig. 1A). For example, ionizing radiation often produces phosphate or phosphoglycolate moieties at the 3’ terminus of DNA and hydroxyl groups at the 5’ end, both of which must be enzymatically processed prior to ligation [6, 7]. Clastogens that cause DSBs through topoisomerase inhibition form phosphotyrosine adducts that can block either DNA end [8]. Loss of nucleotides resulting in 3’ and 5’ overhangs at DSB termini is possible during the formation of the DSB or repair related nucleolytic processing that follows [9], necessitating DNA polymerase activity to produce ligation competent ends. In spite of the complicated processes involved in their repair, the frequency of DSBs in nondividing cells is estimated to be about 10 per day[10], illustrating how robust of the mechanisms evolved to combat are.

Figure 1. Causes of DSBs and Methods of Their Repair.

A. Depictions of the end configurations produced by several different causes of DSBs, illustrating the variety of termini DSB repair pathways encounter. B. Broad overview of the three main pathways of DSBs, showing the major DNA related events. NHEJ directly bridges the two free DNA ends, while both HR and alt-EJ rely on an initial resection step. C. A model of the major events of NHEJ beginning with Ku recruitment to the DSB, followed by formation of the DNA-PK complex, and subsequent recruitment of the NHEJ scaffolding factors allowing for end-bridging. DNA-PKcs autophosphorylation triggers a conformational change that results in eviction of the kinase and short-range synapsis of the two DNA ends allowing for tandem ligation, followed by disassociation of the complex.

The major events that characterize the three identified pathways of DSB repair, Non-Homologous End Joining (NHEJ), Homologous Recombination (HR), and Alternative End Joining (alt-EJ), reflect the two major solutions cells have developed to both bridge the damaged DNA ends and to cope with the diversity of DSBs produced (Fig. 1B). In NHEJ, both damaged ends are directly bridged through protein-protein interactions after detection by repair factors, followed by a variety of possible enzymatic end-processing reactions that produce two ends competent for ligation [5, 11, 12]. In both HR and alt-EJ, recognition of the damaged ends is followed by a 5’−3’ resection step that leaves a long 3’ single stranded overhang [13]. In HR, this overhang anneals to a stretch of corresponding DNA on the sister chromatid or, less frequently, a homologous chromosome [14], anchoring the damaged end to downstream DNA, while alt-EJ accomplishes end-bridging by annealing short stretches of microhomology on the other end of the DSB [15]. The initial resection step in these pathways also serves as an end-processing mechanism by removing any structures that chemically block the DNA termini, largely bypassing the need for specialized end-processing steps.

Outside of S and G2 phase, NHEJ is the major pathway used to correct spurious DSBs [13], and it is also responsible for repairing the programmed DSBs produced during V(D)J recombination [16]. Identification and investigation of the various components in NHEJ has been ongoing for several decades [17], but due to the size and complexity of the cellular machinery involved, structural evidence elucidating the molecular mechanisms that govern repair through this pathway has been difficult to obtain. However, advances in cryo-Electron Microscopy (cryo-EM) over the last decade [18] have made structurally characterizing large protein complexes much more feasible, allowing for the visualization of several key steps of the NHEJ mechanism [19–29]. Combining these structural data with recent findings from other modern biophysical and biochemical techniques has led to remarkable advances in the understanding of the molecular events that occur during NHEJ. As outlined in Fig. 1C a DSB is initially recognized by the Ku70/80 heterodimer, which binds to the damaged DNA end and is responsible for the recruitment of several other NHEJ factors. The second NHEJ factor to arrive at the break is the DNA Dependent Protein Kinase Catalytic Subunit (DNA-PKcs), which pushes Ku farther upstream of the break, replacing it at the DNA terminus [22, 23]. Next, the NHEJ scaffolding factors, X-ray repair cross complementing protein 4 (XRCC4), XRCC4-like factor (XLF), paralog of XRCC4 and XLF (PAXX), and DNA ligase IV (LigIV), are recruited to the complex, allowing for end-bridging to proceed using several protein-protein contacts [19, 20, 26, 27]. The two opposing copies of DNA-PKcs on each side of the DSB undergo autophosphorylation in trans leading to a rearrangement of the complex [19], that ultimately leads to the eviction of DNA-PKcs from the complex, at which point the two DNA ends become accessible for end-processing and ligation [26]. Here we examine the molecular details of these events as revealed through key structural studies of NHEJ, informed by the many elegant biochemical and biophysical investigations conducted in parallel. We also discuss the ways in which our understanding of NHEJ is still relatively immature at this level of mechanistic detail, suggesting new avenues for inquiry and discovery.

2. Initial Recruitment to Double Strand Breaks

After DNA has been damaged, the lesion must be recognized by the cell and repair factors recruited to the site of damage before repair occurs. There are numerous events associated with the cellular recognition of DSBs and initiation of repair including damage specific chromatin modifications[30], accumulation of a multitude of early damage response factors [31–33], damage specific posttranslational modifications of related proteins [34], and even the formation of different liquid-phase condensates that are speculated to sequester the damaged DNA into regions of high local concentrations of repair factors [35, 36]. Precisely how NHEJ functionally relates to these processes is an active area of investigation, but it is generally accepted that the first component of the NHEJ pathway recruited to DSBs is the Ku70/80 heterodimer, likely due to its incredibly high affinity for duplex DNA ends (KD $~$ 2.4 – 0.5 nM) [37, 38] and abundance in the nucleus, which is estimated to be around 500,000 copies per nucleus [39]. Ku70 and Ku80 are two evolutionarily related proteins [40] with a shared domain architecture (Fig 2A) comprising N-terminal von Willebrand (vWA) domain and that flanks the ring-like core and arm domains, together making up the bulk of the protein[41]. The more flexible C-terminal regions (CTR) contain small accessory domains; the Ku80 CTR includes a helical bundle [42, 43] observed to interact with DNA-PKcs in several recent structures[24, 26–28], and Ku70’s CTR contains a SAP domain of indeterminate function that exhibits weak binding to DNA binding [44].

Figure 2. Domain architecture and structural details of DNA-PK complex components.

A. Domain architecture of the Ku80 and Ku70 proteins. Folded domains are indicated by rounded boxes, and flexible regions are represented by bold lines. Confirmed interactions with other NHEJ factors are shown below, indicated by rectangles that follow the same color scheme as the model for NHEJ in Fig. 1C, where applicable. All components are represented roughly to scale. B. Domain architecture of DNA-PKcs following the same conventions as Fig. 2A where applicable. Due to size and complexity of the protein, major domains are designated at the top, with finer structural details shown below. Confirmed contacts with other NHEJ factors are also shown. C. Structures of Ku70/80 (1JEY) and inactive DNA-PK complex III (7K1J) bound to DNA. DNA footprints of both proteins are shown, and arrow designates pushing of Ku70/80 along the DNA away from the termini by DNA-PKcs. D. Structural rearrangements observed during DNA-PKcs activation with the protein colored in the same color scheme as Fig. 2B. The rotation of the PRD out from the catalytic cleft allows proper alignment of the P-, activation, and catalytic loops and closure of the domain. The HhH motif that binds DNA in the protein’s string region is pulled towards the nucleic acid causing a disorder to order transition of the flanking loops. Activation of DNA-PKcs also leads to the stabilization of the DEB helix which binds directly to the DNA terminus.

Ku’s role in NHEJ is to bind to the DSB break, protecting it from extraneous enzymatic activity and to recruit other NHEJ factors to the site of damage. The Ku heterodimer binds to DNA in an idiosyncratic manner by threading onto the DNA[45], making it ideal for sensing and binding to DSBs. This propensity for free DNA ends is remarkably flexible, with early in vitro biochemical studies showing that Ku is capable of binding to a wide variety of DNA termini including those with long 3’ and 5’ overhangs, hairpin DNA, various configurations of terminal phosphate groups [37, 38, 45]. The crystal structure of Ku bound to DNA reveals it occupies a 14 bp stretch of DNA immediately flanking the DNA termini [41]. Interestingly, it appears that Ku threads onto DNA in a specific orientation, with the core of Ku70 closer to the DNA end, a feature that is essential for proper recruitment of downstream NHEJ factors but has not been fully explained by the structural data. Once bound to DNA, Ku serves as the major nucleating factor for the rest of the core NHEJ machinery, forming structurally characterized contacts with the factors DNA-PKcs [22–24], XLF [46], DNA Ligase-IV [26, 27], and PAXX [19, 20]. Several accessory NHEJ factors are also recruited by Ku, notably aprataxin and PNKP like factor (APLF), modulator of retrovirus infection (MRI), and Werner syndrome helicase (WRN), which all compete for the same pocket on the surface of the Ku80 vWA domain via their Ku binding motifs (KBMs) [47, 48]. The NHEJ specific polymerases pol $\text{µ}$ and pol $\text{λ}$ also make reported interactions with Ku through their BRCT domains, which appear to be indispensable for stimulating their activity on DNA in NHEJ contexts [49]. By binding directly to the DSB termini and recruiting downstream NHEJ factors, Ku plays a major role in the initiation of the NHEJ repair pathway.

3. Formation and Activation of the DNA-PK Complex

The next NHEJ factor to be recruited is to the DSB is DNA-PKcs, a member of the phosphatidylinositol 3-kinase-related kinase (PIKK) family, which together with Ku70/80 and DNA forms the DNA-PK complex. DNA-PKcs is a large, 469 kDa single-chain protein with a complicated domain architecture (Fig. 2B) comprising N-terminal DNA binding N-HEAT region, followed by an M-HEAT region that forms a solenoidal cradle encircling the DNA end. The C-terminal FAT-Kinase domains of DNA-PKcs form the head region, which contains the ATP-binding catalytic pocket and several sites of reported protein-protein interactions [50]. In addition to monomeric DNA-PKcs, a dimeric form has been observed that formed through significant contacts between N-heat, M-heat, and head regions [19]. This state resembles the dimers of the other repair related PIKK kinases ataxia-telangiectasia mutated (ATM) [51, 52] and ATM and RAD3-related (ATR) [53]. The purpose of dimerization is unclear, but it has been suggested that it may play a role in the induction of DSB repair by sensing oxidative stress [54], or serving as a reservoir for the repair factors that disassociate into the active monomeric form [19].

The formation and activation of the DNA-PK complex is a multistep process, that involves a number of distinct conformational changes to DNA-PKcs. Many of these intermediate states have been directly visualized by cryo-EM [22–24, 29], owing in large part to the large size and stability of DNA-PKcs, making it amenable to the technique. Although DNA-PKcs can bind DNA in the absence of Ku under low-salt conditions, it has a much higher affinity to Ku-DNA (KD $~$ 35 pM) [55], and its arrangement on the DNA end distal to Ku indicates that it almost certainly binds to the DSB after Ku recruitment. Upon binding, DNA-PKcs pushes Ku along the DNA away from the DNA end causing it to translocate along the double helix about 15 bp (Fig. 2C), accommodating a footprint of 28–29 bp for the entire DNA-PK complex [22, 23], although DNA as short as 26 bp has been reported to be sufficient for the formation of an active DNA-PK complex [56]. Once assembled, DNA-PK protects the DNA ends from further processing [57], setting the stage for the recruitment of the NHEJ scaffolding factors and synapsis.

The maturation of DNA-PK from an inactive end-protecting state to an catalytically active state that supports end-processing and downstream NHEJ events involves a number of different conformational rearrangements (Fig. 2D), primarily of DNA-PKcs. As its name implies, DNA-PKcs kinase activity is dependent on its binding to DNA, which is further stimulated when it binds to Ku-DNA [58]. Structures of activated and intermediate states of DNA-PK have revealed that a DNA interacting helix-hairpin-helix (HhH) motif (residues 813–836) located in the N-HEAT domain and loops immediately flanking it are stabilized and become ordered when activated [24, 25]. The resulting compaction of the N-terminal region of DNA-PKcs may contribute to the observed rigid-body rotation of the FAT-kinase domain relative to the rest of the protein. These rearrangements coincide with the appearance of the DNA end binding (DEB) helix (residues 2736–2767), which inserts itself between the terminal base pair of the DNA termini, occluding further access to the DNA. Several specific changes to the DNA-PKcs kinase domain are responsible for its activation. In its inactive form, the PIKK regulatory domain (PRD, residues 4009–4039) sits inside the catalytic cleft, inhibiting kinase activity [22, 23, 50]. Upon activation, the PRD has been observed to rotate outwards by 115° [24] or become disordered [25], allowing the active site to bind to ATP accommodated by interactions with the P-loop (3729–3735), activation loop (3940–3963), and catalytic loop (3919–3927). It is important to note that while most of the information regarding these conformational changes comes from DNA-PK structures, many of these events may also occur after the formation of the larger synaptic NHEJ complexes when both strands of DNA have been brought into proximity, and DNA-PKcs makes extensive dimeric contacts. Likewise, several of the phosphorylation dependent conformational changes discussed in the following sections are also observed in the DNA-PK structures, which ostensibly represent the earlier steps of NHEJ. Delineating precisely when during DSB repair these DNA-PKcs conformational changes occur will require further investigation, and it is likely that the kinase transitions dynamically between conformations during the process.

4. End-bridging

As discussed above, one of the unique challenges posed by DSBs is the fact that the DNA up and down stream of the site of damage becomes physically separated in the cell and must be brought back in to proximity before repair. Because it lacks the homology directed mechanisms utilized by other DSB pathways, exactly how NHEJ effectively bridges the two ends of a DSB has been one of the most important questions in the field. Despite several key studies elucidating these mechanisms [26, 27, 59, 60] there are still several competing models that need to be reconciled. These can broadly be grouped into two categories; those involving DNA-PKcs and those where DNA-PKcs is absent, which we discuss in turn. Further work is needed to fully understand the requirement for DNA-PKcs in repair and to incorporate the many disparate proposed models for end-bridging. As we discuss below, it is likely that cells rely on several redundant end-bridging mechanisms that can compensate for one another so that DSB repair remains robust in a variety of circumstances.

4.1. DNA-PKcs Mediated End-bridging

Based on the structures of the LR synaptic complexes [19, 20, 26, 27] end-bridging is mediated by a combination of protein-protein interactions that hold the two DNA ends in proximity to one another but not in a position that allows for ligation. This long-range arrangement was first predicted by single-molecule FRET experiments conducted in Xenopus laevis egg extracts that identified a state in which the two ends tethered together, but at a distance greater than 100 Å [59], which was later confirmed to be 115 Å by cryo-EM [26, 27]. As is evident from the structure of the LR complex (Fig. 3A), the two opposing copies of DNA-PKcs make several contacts with one another, the most prominent being two loops (896–903 and 946–950) in the forehead region interacting with the YRPD interacting loop (2569–2585) on the opposite protomer [26, 27]. The YRPD interacting loop and the corresponding YRPD motif are both highly conserved regions of DNA-PKcs [61], and their participation in this interface points to a possible role in sensing the close juxtaposition of two copies of DNA-PKcs, signaling successful end-bridging. DNA-PKcs in the LR complex adopts an activated conformation, with the DEB helix clearly discernable and kinase domain closed with an evicted PRD.

Figure 3. NHEJ scaffolding factors and end-bridging in the LR complex.

A. Structural model of the PAXX mediated LR complex (8EZA). Subunits are colored as in Fig. 1C. B. Structures of the PAXX, XRCC4, and XLF highlighting the similarities in their structures. XLF and XRCC4 are shown in their arrangement captured in in the PAXX mediated LR complex (8EZA), while the crystal structure of PAXX is shown due the inability to visualize its head regions in the complex. C. Domain architecture of the NHEJ scaffolding factors XRCC4, XLF, and PAXX, figures follow the same conventions as shown previously. The CK2 phosphorylation site on XRCC4 is indicated with a red line. D. Comparison of the DNA bound Ku70/80 (1JEY) to its PAXX and XLF bound state seen in the PAXX mediated LR complex (8EZA). Arrows depict the rotation of both vWA domains outwards from the core of the heterodimer to accommodate the binding of both PAXX and XLF KBMs. E. Top view of both copies of DNA-PKcs from the PAXX mediated LR complex (8EZA) highlighting the proximity of each protomer’s kinase domain (cartoon model with tube helices) and the ABCDE cluster (shown in red) on the opposite DNA-PKcs. This distance is measured to be $~$60 Å [26].

In addition to DNA-PK, other core NHEJ factors are responsible for end-bridging; LigIV, XRCC4, XLF, and PAXX [62]. The latter three of these proteins have a similar architecture (Figs. 3B, 3C); an N-terminal head domain followed by a coiled-coiled domain that mediates their homodimerization and a disordered C-terminal tail [63–65]. Despite their structural similarity, each of these scaffolding factors has a distinct set of interactions with other NHEJ factors. One of the most striking of these interactions occurs between XRCC4 and XLF, mediated by a pseudo symmetric beta-sheet zipper of hydrogen bonds between their two head domains combined with the XLF Leu-115 inserted into a hydrophobic pocket of XRCC4 that forms the XRCC4-XLF scaffold [66, 67]. The coiled-coil region of the XRCC4 homodimer also forms an stable complex with the two BRCT domains of LigIV (termed X4L4), which flexibly tethers the ligase to the core complex, allowing its catalytic domain to dynamically access the DNA ends [68]. One of these BRCT domains also interacts with the arm of Ku70, serving as a major anchoring point of X4L4-XLF to DNA-PK [26, 27]. Additionally, the C-terminal tail of XRCC4 is phosphorylated at T233 by CK2, which recruits the fork-head associated (FHA) domains of the accessory NHEJ proteins PNKP [69, 70], APLF, and APTX [71].

While XRCC4 has a unique role in positioning LigIV proximal to the DSB and indirectly engaging with Ku, there is growing evidence that the functions of XLF and PAXX may have some overlap. XLF was originally identified as a component of the NHEJ pathway in 2006 [72, 73], but subsequent characterization revealed the effects of its ablation were extremely variable and dependent on the biological context [74]. PAXX, discovered almost a decade later in 2015 [65, 75, 76], has since been shown to possess several of the same functional attributes of XLF, and its presence in the experimental systems used to investigate XLF may be responsible for some of the confounding results. For example, smFRET experiments conducted on Xenopus laevis egg extracts showed that extracts from XLF deficient eggs were still capable of synapsis[59], an observation that could be explained by the presence of PAXX in the reactions. Using a purified in vitro system, single molecule studies using magnetic tweezers have confirmed that DNA-PK and PAXX are sufficient for end-bridging[60]. Furthermore, while the first structures of the NHEJ synaptic complexes included only the XLF and XRCC4 scaffolding factors [26, 27], more recent structural work has revealed the striking similarities in which XLF and PAXX interact with Ku to help mediate synapsis [19, 20]. XLF and PAXX both harbor KBMs in their respective CTRs that bind to the clefts between the core and vWA domains of Ku80 and Ku70 respectively, and the deletion of either of these sequences has been shown to negatively affect end-joining under certain circumstances [77, 78]. These peptides are accommodated by the rotation of both vWA domains of Ku70/80 outward, causing the protein to adopt an ‘open’ conformation (Fig. 3D). Additionally, DNA-PK dimers mediated only by PAXX were also reported in both structural studies, consistent with PAXX having the ability to initiate end-bridging independently to X4L4-XLF recruitment, in line with data from the single molecule studies. However, from in vitro ligation assays, it does not appear that PAXX alone can support a fully functional synaptic complex that is capable of ligation [19], indicating that downstream NHEJ enzymatic activity requires further stabilization by XLF.

Another synaptic complex that bridges the two DNA ends only through DNA-PK dimer interactions without the use of the scaffolding factors XRCC4, XLF, and PAXX has been captured using cryo-EM [28]. The structure of this dimeric DNA-PK complex shows that dimerization occurs through interactions between the M-HEAT cradle region of DNA-PKcs, and are additionally stabilized by C-terminal alpha-helix of Ku80 binding in trans to the opposing DNA-PKcs near this interface. Mutations on the surface of this DNA-PKcs interface interfere with nucleolytic processing of DNA ends, but they do not have the same inhibitory effect on end-joining as mutating the LR complex DNA-PKcs dimerization interfaces [79]. Single molecule studies also indicate that DNA-PK alone is not sufficient for synapsis, even in purified systems [60]. The purpose for this alternative dimeric form of DNA-PK is still an active line of inquiry, and it is unclear if it is an intermediate of routine NHEJ.

Taken together, the structural and biophysical evidence supports a model of end-bridging mediated through a variety of different protein-protein interactions. There is an apparent redundancy between some of these interactions, which may confer flexibility in the order in which NHEJ components are assembled as well as compensating for any deficiencies in specific factors or weakening mutations. For instance, a recent in vivo study showed that dimerization of DNA-PKcs in the LR complex compensated for mutations that weakened the ability of XLF to form a stable homodimer [80]. Additional factors that have not been integrated into this canonical end-bridging mechanism are also likely to reveal further functional redundancy in this pathway.

The relative abundance of DNA-PKcs in the nucleus [81], its rapid localization to sites of irradiation [82, 83], and extensive structural and functional interactions with different NHEJ factors suggest that it plays more than just an accessory role in bona fide NHEJ. One particularly illustrative line of evidence comes from the comparisons of mice deficient in DNA-PKcs to those harboring catalytically dead mutants of the kinase [84]. Strikingly, far greater genomic instability and defects in V(D)J recombination were observed in the DNA-PKcs mutant mice compared to those lacking the kinase, which argues that the redundancy inherent to the molecular mechanisms of NHEJ are able to compensate for the lack of DNA-PKcs but that an inactive form will cause repair to stall after the formation the DNA-PK or LR complex. Another important observation is that DNA-PKcs is not represented in prokaryotes and conspicuously absent in many model organisms including eukaryotes such as S. cerevisiae and D. melanogaster [61], which has given an impression that it is a more recently evolved DNA repair factor. However, more recent phylogenetic analysis has identified DNA-PKcs presence in a broad range of eukaryotes including fungi, plants, and protists [61]. This indicates two important points; first that DNA-PKcs is well-represented across many species, ostensibly participating in NHEJ in a diverse set of systems, and second that, despite the kinase’s prevalence, redundant mechanisms capable of compensating for its loss are readily available to organisms. For these reasons, we speculate that DNA-PKcs participates in repair wherever it is present, despite the fact that it is not strictly required for NHEJ.

4.2. Alternative Modes of End-bridging

There are several proposed alternative mechanisms explaining end-bridging in NHEJ that incorporate different combinations and arrangements of the core NHEJ factors Ku, and X4L4-XLF. Perhaps the most widely investigated of these derives from the observation that the X4L4-XLF scaffold and Ku alone are sufficient to mediate synapsis. Filaments composed of repeating units of XRCC4-XLF readily form in vitro and have been speculated to wrap around the DNA flanking the DSB to hold both ends together. These filaments have been characterized structurally [66, 67] and biophysically using single molecule techniques [85, 86], and there is even evidence that they are represented to some extent in vivo [46]. Furthermore, in vitro reconstitution of the ligation reaction from purified NHEJ proteins proceeds readily in the absence of DNA-PKcs, while remaining dependent on Ku, XLF, and X4L4 even on substrates that are too short to bind an X4L4-XLF filament [26, 87, 88]. Other studies in purified systems have reported that a more flexible complex comprising only Ku70/80 and X4L4 tether the DNA ends together potentially representing the first step in end-bridging prior to recruitment of XLF and the transition to the SR synaptic complex [89, 90].

In addition to the core NHEJ factors, other accessory proteins have been identified to play a role in end-bridging. APTX and PNKP Like Factor (APLF), a largely disordered protein that contains an FHA domain known to interact with the phosphorylated tail of XRCC4 [71] and a KBM that binds to an outwards facing cleft of the Ku80 vWA [46] may contribute to assembly of the complex [91]. Single molecule experiments using magnetic tweezers show that APLF in combination with Long Noncoding RNA NIHCOLE, a long noncoding RNA previously discovered in cancer cells to promote end-joining [92], is sufficient for forming stable synapsis of Ku bound DNA ends [93]. Polymerase $\text{µ}$ has also been shown, via single molecule techniques, to mediate synapsis of Ku and X4L4 bound DNA that contain 3’ overhangs of at least 1 nt [89].

Finally, it is even possible that NHEJ borrows factors that are central to other DSB repair pathways to serve end-bridging functions. The Mre11-Rad50-Xrs2/Nbs1 complex (MRX/N), responsible for resection in HR repair, has been shown to be indispensable end-bridging in S. cerevisiae, an organism lacking DNA-PKcs, by coordinating with nonhomologous end-joining protein 1 (Nej1), a ortholog of XLF [94]. The ability for two copies of Rad50 to associate with one another in trans via a Zn²⁺ “hook” near the end of its long coiled-coil arm suggests a potential mechanism for the complex to tether both damaged ends [95]. Many functional and physical interactions between MRX/N and NHEJ factors, including Ku70/80 and DNA-PKcs, have been uncovered [96], and understanding the coordination between these proteins will lead to important insights into both NHEJ and HR, as well as the processes governing the choices between them. Given the multivalency of the interactions formed by the different NHEJ factors, it is unsurprising that there could be multiple mechanisms that give rise to end-bridging in NHEJ, and future work will be needed to clarify the relations between them.

5. Transition to Synapsis

Due to the distance the DNA ends are held apart in the LR complex and the manner in which DNA-PKcs protects the termini, occluding access with its DEB helix, there are several structural rearrangements to the complex that occur before NHEJ can continue. As alluded to earlier, DNA-PKcs kinase activity is necessary for proper end-joining both in vivo [97, 98] and in purified systems that include DNA-PKcs [26]. DNA-PKcs’s phosphorylation targets in vitro are myriad including Ku70 and Ku80 [99], PNKP [100], the NHEJ polymerases $\text{µ}$ and $\text{λ}$ [101], and it is capable of phosphorylating itself in both cis and trans. The current purpose of the majority of these posttranslational modifications is unclear, but DNA-PKcs autophosphorylation, the most common target of its kinase activity [86], is proving to be crucial for end-joining. Two highly conserved regions of DNA-PKcs autophosphorylation have been identified, the ABCDE cluster comprising several serines and threonines followed by a glutamine residue (T2609, S2612, T2620, S2624, T2638, T2647) [97] and the PQR cluster (S2023, S2029, S2041, S2053, S2056) [98]. Phosphoablation (S or T to A) mutations of both sensitize cells to radiation, implicating their importance in DNA repair. The structures of the LR complex reveal that the kinase domain of each DNA-PKcs is juxtaposed with the ABCDE and PQR clusters of the opposing DNA-PKcs (Fig. 3E) [26, 27]. While both of these loops are too flexible to be resolved in the structures, their rough location could be estimated to be 60 Å and 75 Å respectively from the kinase active site, leading to speculation that DNA-PKcs autophosphorylation in trans at these sites would be able to cause a conformational change driven by the resulting electrostatic repulsion of the phosphate groups that would ultimately lead to full synapsis.

Recently, by supplying ATP a LR complex stabilized by PAXX, the structure of an intermediate autophosphorylated intermediate was captured, providing structural evidence for this idea of autophosphorylation triggered state transition (Fig. 4A) [19]. A new DNA-PKcs dimerization interface comprising several regions between residues $~$1700–1930 appears, indicating DNA-PKcs may still play an end-bridging role at this stage. It is worth noting that this region shares some overlap with the dimerization interface present in the Ku80 CTR mediated DNA-PK dimer [28]. Strikingly, the ABCDE loop becomes ordered in the ATP-state, and it is observed to occlude a large portion of the dimerization interface present in the LR complex (Fig. 4B). There are discernable changes in the YRPD interacting motif, and the DEB helix becomes disordered in this state, ostensibly allowing access to the DNA ends, potentially by any available end-processing enzymes that have been recruited to the complex. The DNA-PKcs cradle region also adopts a more open conformation, that could also assist in accommodating end-processing enzymes (Fig. 4C). It is intriguing that remarkably similar structural changes are observed in the recently determined non-synaptic DNA-PK-Artemis complex where these rearrangements are described as a mechanism by which DNA-PKcs can autoinhibit itself while end-processing is carried out by the Artemis nuclease [29]. In the LR-ATP state, DNA-PKcs protomers undergo a profound conformational rearrangement rotating roughly 60° in the plane of DNA, which likely contributes to its subsequent eviction from the complex (Fig. 4D). These structural data provide strong evidence for a mechanism in which DNA-PKcs autophosphorylation as the key event governing NHEJ state transition from long-range to short-range synapsis.

Figure 4. State transition from long-range to short-range synapsis.

A. Structural model of the LR-ATP state. B. Comparison of the surface of DNA-PKcs from both LR and LR-ATP complexes highlighting the conformational changes observed in the YRPD and YRPD-interacting motifs (yellow) and the stabilization of the ABCDE phosphorylation cluster (red). The region shown participates in dimerization of DNA-PKcs in the LR complex C. Side view of DNA-PKcs showcasing the ATP induced opening DNA-PKcs. The N-HEAT region from the LR complex is shown in grey, LR-ATP state is colored. Directly below is the structural model of DNA-PKcs bound to Artemis (7SGL) from the recently reported DNA-PK-Artemis complex, which exhibits a similar conformational change that accommodates the nuclease’s catalytic domain. D. Comparison of the LR and LR-ATP orientations of DNA-PKcs relative to the scaffold highlighting the rotation it undergoes upon autophosphorylation in trans. E. Structural model of the SR synaptic complex.

6. End-processing

Owing to the diversity of damaging agents that can give rise to DSBs, NHEJ must be capable of joining DNA with a variety of different end configurations. The process by which incompatible DNA termini are enzymatically converted to be competent for ligation involves a wide variety DNA repair enzymes that can be roughly grouped into three categories based on their activity. First are the enzymes that directly resolve chemical adducts that blocking ligation such as PNKP [102], APTX [5], sSecond are the NHEJ specific polymerases pol $\text{µ}$, pol $\text{λ}$, and TdT fill short gaps and overhangs of up to about 5 nt in length[12, 103]. Finally, the nuclease Artemis has been identified to process DNA ends through cleavage of overhangs, limited resection, and endonucleoylytic activity[104]. The nuclease Mre11 [105] and helicase/exonuclease WRN [106] have also been shown to interact with the NHEJ machinery and could potentially serve a similar role. Most of the structural details of how these factors coordinate with the core NHEJ factors have not yet been revealed, and it is still unclear at which steps of the repair process the complexes are capable of supporting end-processing. However, some insight can be derived from what is known in about each of these factors individually in combination with the structures of the synaptic NHEJ complexes.

One of the key functions of the core NHEJ machinery is to protect the free DNA ends from erroneous end-joining or aberrant enzymatic activity. Ku occupying the terminal stretch of DNA sterically blocks the recruitment of other DNA binding proteins that would typically have activity at free DNA ends, such as DNA ligases [37], and, as detailed above, DNA-PKcs also provides protection of these ends after it binds to Ku-DNA. At some stage of NHEJ, this protection must be at least temporarily relieved to accommodate the end-processing factors. One possibility is that the conformational rearrangements associated with DNA-PKcs autophosphorylation create a large enough opening in the M-HEAT regions of the protein and a sufficient amount of DNA for end-processing to occur. This seems to be the case for Artemis, which can bind directly to a DNA substrate in complex with an activated DNA-PK [29], favoring a model of early end-processing by the nuclease. The C-terminal tail of Artemis has been a well-documented target of phosphorylation both in vitro and in vivo, with some studies suggesting that DNA-PKcs kinase activity at these sites being important for activation of the nuclease [107–109] and others identifying residues phosphorylated by ATM with that modulate its activity [110, 111]. It has also been shown that DNA-PKcs autophosphorylation is required for Artemis activity [112], which agrees well with the structure Artemis-DNA-PK complex [29]. Furthermore, the structure of Artemis-DNA-PK shows that the Artemis CTR forms an extended surface with the FAT and M-HEAT regions of DNA-PKcs, confirming decades of biochemical evidence for this interaction [21, 29].

There are several features that separate the other end-processing factors from Artemis. First, some of end-processing factors including the NHEJ polymerases and PNKP are reported targets of DNA-PKcs and ATM phosphorylation, but this posttranslational modification has not been observed to affect their activity [100, 101]. Also, in contrast to Artemis which makes extensive contacts to the core NHEJ machinery, current biochemical evidence indicates that the other end-processing factors are recruited to the core NHEJ machinery primarily through relatively small accessory domains - the polymerases via a BRCT domain to Ku interaction [49, 113] and PNKP, APLF, and APTX via FHA domain to XRCC4 interaction [71]. Notably, these interactions are with NHEJ factors that are present both before and after DNA-PKcs is evicted from the complex, leading to the possibility that end-processing by these enzymes may proceed at a different step in repair than Artemis. Single molecule experiments also provide direct evidence that NHEJ polymerase and Tdp1 activity occurs during short-range synapsis [114]. It is also worth noting that the existing crystal structures of the catalytic domains of these enzymes bound to DNA substrates almost invariably show the protein bound to DNA both upstream and downstream of the DSB [115–120], only possible when the DNA ends are unprotected and capable of short-range synapsis. All of these observations support to a model of NHEJ in which most end-processing events occur after DNA-PKcs has been evicted from the complex, while certain activities, such as nucleolytic cleavage by Artemis, may be supported at different steps of end-joining.

7. Tandem Ligation and Complex Dissolution

Current models suggest that after DNA-PKcs autophosphorylation, it is evicted from the complex allowing for the bridged DNA ends to fully synapse, forming the SR synaptic complex (Fig. 4E). The lower resolution reported in the structure of this state (8.4 Å) [26] is indicative of the complex’s inherent flexibility compared to the various LR complexes held together through DNA-PKcs interactions. A single catalytic domain of LigIV bound to the DSB junction is observed, while the BRCT domains of two copies of the ligase anchored to their respective protomers of XRCC4 are discernable, denoting the presence of two LigIV proteins in the complex. Unlike most other DNA ligases, LigIV is a single-turnover enzyme [121, 122] explaining the requirement for both copies, which tandemly ligate both strands of DNA. Several helices on the surface of the LigIV DBD interact with the vWA of Ku70, stabilizing the catalytic domain at the site of the break. As is observed in the crystal structures of the LigIV catalytic core, the enzyme binds to roughly 18 bp of DNA flanking the DSB and adopts a closed conformation held together by a molecular ‘latch’ between its OBD and DBD regions[123]. In addition to LigIV, there is evidence that the activity of other NHEJ enzymes may be supported at this stage[114]. This makes sense, given the accessibility of the fully synapsed DNA ends in the SR complex, and further investigation on this matter is needed to uncover the molecular mechanisms driving this putative iterative process of end-processing and ligation.

After ligation, most of the NHEJ factors are able to freely disassociate from the repaired DNA. However, the Ku heterodimer remains sterically trapped on the DNA due to its closed ring-like structure and the manner in which it binds DNA. Exactly how Ku is removed from DNA remains unclear, but there are a multiple explanations involving both proteolysis or temporary cleavage of the DNA[124, 125]. After DNA repair, Ku is postranslationally modified by the addition of NEDD8, a ubiquitin-like protein, to the Ku80 protomer, which recruits the p97 complex, a ubiquitin-directed unfoldase. This mechanism appears to be capable of freeing least Ku80 from DNA, which would likely result in Ku70 degradation either directly or indirectly. Alternatively, it has been proposed that the repair related nuclease Mre11 can nick DNA near Ku, allowing for it to be pushed off of the DNA end [126], although this mechanism may be more relevant in cases where Ku remains bound to DNA after another repair pathway such as HR has already outcompeted NHEJ and initiated repair. The lack of understanding of these issues illustrate that there are still several aspects of NHEJ termination that remain inadequately explained.

8. Conclusions and Future Directions

As the structural biology of NHEJ matures, the molecular mechanisms it uses to carry out and regulate DSB repair will continue to be elucidated. The cryo-EM investigations of NHEJ complexes carried out over the past few years have yielded an increasingly detailed picture of how the NHEJ factors coordinate their activities to identify and protect free DNA ends, facilitate the localization of both ends to one another, and support reactions required to process and ligate the damaged ends. The complex role of DNA-PKcs in this process has been increasingly explained through structural data. As more orthogonal evidence is integrated into these models, nuances regarding the specific order of events in NHEJ and activities of the various appendix NHEJ factors will undoubtedly be clarified.

Looking beyond these immediate concerns, there are still many looming fundamental questions regarding NHEJ, as well as DSB repair in general. One striking weakness of the field is the relative dearth of information regarding DSB repair in chromatin. Although some progress has been made, including the identification of specific NHEJ factors that act as histone chaperones [127, 128], models of NHEJ will not be comprehensive until they include the mechanisms by which nucleosome bound DNA is repaired. There are also outstanding questions regarding the function and inhibition of the NHEJ machinery at telomeres, which evade erroneous DSB repair to prevent chromosomal fusions yet recruit some components of the NHEJ machinery [129, 130]. Finally, the fundamental processes that govern the regulation of the different DSB repair pathways relative to one another have yet to be fully explicated. Beyond profoundly advancing our understanding of DNA repair, solving the questions related to pathway choice has the potential to dramatically impact human health through its effects on cancer research and its applications in enhancing CRISPR-Cas biotechnologies.

Highlights:

• We summarize findings from the recent cryo-EM structures of synaptic and presynaptic NHEJ complexes.

• Structures of DNA-PKcs coordinating with the other NHEJ factors suggest molecular mechanisms underlying its roles in bridging broken DNA ends and regulating DSB repair.

• Model for NHEJ mediated DSB repair through a succession of dynamically assembled protein complexes.