In order for LigD to participate in NHEJ, it requires its binding partner, Ku. The prokaryotic Ku protein is a homodimer of approximately 30–40 kDa, consisting of a core domain conserved with the eukaryotic Ku homolog, and a C-terminus unique to bacteria (McGovern et al., 2016). The C-terminus of Ku can be further subdivided into a minimal and an extended region. The minimal C-terminal region is conserved amongst bacteria, while the extended C-terminal region has low sequence conservation and is highly variable in length between organisms (Figures 2A,B) (Kushwaha and Grove, 2013b; McGovern et al., 2016). The core domain is predicted to form a ring-shaped structure that encircles DNA, much like eukaryotic Ku70/80, based on high sequence homology, and various in silico atomic structures from B. subtilis and M. tuberculosis (McGovern et al., 2016; Jumper et al., 2021; Öz et al., 2021). The structure of the C-terminus, though, is likely variable. Disorder predictions indicate the C-terminus is an intrinsically disordered region (Oates et al., 2013), which correlates with findings from small-angle x-ray scattering studies of B. subtilis Ku (McGovern et al., 2016). We carried out in silico modelling of Ku homodimers using the ColabFold notebook, which uses the AlphaFold algorithm (Jumper et al., 2021; Mirdita et al., 2021). The resulting predictions of Ku shows that the shorter C-terminus of M. tuberculosis Ku may take on some structure, with alpha helices binding the C-terminus to the core, interspersed with disordered loops (Figure 2C). P. aeruginosa and B. subtilis Ku, though, have longer, disordered C-termini, corroborated by ColabFold’s low confidence score in positioning the modelled C-termini (Figures 2D,E) (Jumper et al., 2021; Mirdita et al., 2021; Öz et al., 2021).

Ku recruits LigD to the DSB and stimulates both ligase and polymerase activities of LigD (Aravind and Koonin, 2001; Weller et al., 2002; Della et al., 2004; Zhu and Shuman, 2010; McGovern et al., 2016; Öz et al., 2021). The Ku C-terminus of B. subtilis and P. aeruginosa Ku are credited with stimulating ligation, while the Ku B. subtilis minimal C-terminal region is needed to interact with LigD (Zhu and Shuman, 2010; McGovern et al., 2016; Öz et al., 2021). The extended C-terminal region of B.subtilis and M. smegmatis Ku can bind supercoiled DNA (Kushwaha and Grove, 2013a; McGovern et al., 2016), while the B. subtilis Ku extended C-terminal region also restricts movement along dsDNA (McGovern et al., 2016; Öz et al., 2021). It is likely this combination of interacting with LigD and binding to DNA that permits Ku to tether LigD and then stimulate LigD polymerase and ligase activities at the DSB.

DNA DSBs are rarely a clean break with complementary ends containing a 5′-phosphate and 3′-hydroxyl group. Depending on the DNA damage, there may be gaps in the nucleotide sequence or a phosphate group in place of a hydroxyl at the 3′-end of the DNA (Andres et al., 2015). To repair the DNA to ligation-competent 5′-phosphate and 3′-hydroxyl ends, LigD possesses multiple enzymatic functions. These activities are contained in conserved structural domains that are briefly: the polymerase domain (POL) for the addition of nucleotides; the phosphoesterase/nuclease domain (PE) to convert 3′-phosphate groups to hydroxyl groups; and the ligase domain (LIG) to seal the phosphodiester backbone of the DNA (Weller and Doherty, 2001; Weller et al., 2002; Gong et al., 2005; Pitcher et al., 2005; Zhu and Shuman, 2005a, Zhu and Shuman, 2005b; Zhu and Shuman, 2010; de Vega, 2013). These domains are conserved across bacterial species, however not every domain is always found within a LigD homolog, nor maintained in the same primary structure (Figure 3). For example, B. subtilis LigD lacks the PE domain, with only the LIG domain at the N-terminus and the POL domain at the C-terminus (de Vega, 2013). Meanwhile, M. tuberculosis and P. aeruginosa possess all three domains, but rotate the sequential order of each domain between organisms (Pitcher et al., 2005; Zhu and Shuman, 2005a; Zhu and Shuman, 2005b; Zhu and Shuman, 2006). Given the changes in the sequential order, it remains to be seen how LigD coordinates the processing and ligation of complex DNA end damage in the context of the wild-type protein. Ku may play a role in coordination, as M. tuberculosis Ku directly interacts with the POL domain and weakly interacts with the LIG domain (Della et al., 2004; Pitcher et al., 2005; Wright et al., 2010). However, atomic structures resolved by x-ray crystallography and nuclear magnetic resonance of the POL, PE, and LIG domains in different enzymatic states, combined with in silico predictions of LigD can provide insights into the unique and complementary functions of each domain within LigD. A list of the experimentally determined structures of LigD discussed in this review can be found in Supplementary Table S1.

The LigD LIG domain maintains the conserved ATP-dependent ligation mechanism found in all DNA ligases, which uses three nucleotidyl transfer reactions to seal a DSB (Weller et al., 2002). First, a nucleophilic attack by the catalytic lysine on the α-phosphorus of ATP creates a covalent ligase-AMP intermediate and releases pyrophosphate as a by-product. The resulting configuration leaves the 5′-phosphate at the DNA break site primed to attack the α-phosphorus of the ligase-AMP intermediate. This reaction transfers the AMP moiety onto the 5′ end of the DNA, creating a DNA-adenylylate intermediate. In the final step, the phosphodiester bond is formed after the 3′-hydroxyl attacks the DNA-adenylylate, releasing the AMP and sealing the DNA backbone (Tomkinson et al., 2006). Interestingly, while the ligase mechanism is typically associated with sealing a nick between the ends of deoxyribonucleotides, the LIG domain in P. aeruginosa, M. tuberculosis and Agrobacterium tumefaciens preferentially seals nicks with a 3′-monoribonucleotide (Zhu and Shuman, 2008; Unciuleac et al., 2019).

Current x-ray crystal structures of the LIG domain in the Protein Data Bank are from M. tuberculosis LigD and capture both the pre-adenylylation state (PDB 6NHZ) (Unciuleac et al., 2019) and the covalent LIG-AMP intermediate (PDB 1VS0) (Akey et al., 2006), providing mechanistic insight on the initiation of ligation. The LIG domain is sub-divided into an N-terminal nucleotidyltransferase (NTase) domain and an oligonucleotide binding (OB) domain, which is most similar to human LigI (Figure 4A) (Akey et al., 2006; Unciuleac et al., 2019). The NTase domain forms a nucleotide binding pocket and houses the catalytic lysine that forms the ligase-AMP intermediate (Figures 4A,C) (Akey et al., 2006; Unciuleac et al., 2019). A linker connects the NTase domain to the OB domain, which binds the nicked DNA and positions the DNA into the active site for repair (Akey et al., 2006; Unciuleac et al., 2019)

In the pre-adenylylation state, the OB domain forms a cap over the ATP-binding pocket in the NTase domain (Figure 4A). The linker region between these two allows the OB domain to move and uncover the ATP-binding pocket, as seen in the crystal structure of the LIG-AMP intermediate. The OB domain moves ∼80^°, which increases the space for DNA to bind and be positioned in the active site for repair (Figure 4C) (Unciuleac et al., 2019). We aligned the conserved catalytic core of the human NHEJ ligase LigIV (PDB 6BKG) (Kaminski et al., 2018), bound to a nicked DNA substrate, with the LIG-AMP intermediate structure (RMSD = 2.7 Å). While the fit is not perfect, the alignment shows the 5′-end of the DNA nick within the vicinity of the nucleotide binding pocket, and that opening of the OB domain during catalysis of ligation is likely necessary for active site access (Figure 4B). The movement of the OB domain is also predicted to facilitate formation of the ligase-AMP intermediate and release of the pyrophosphate. In the “closed” state, the OB domain stabilizes the γ-phosphate of ATP and when “opened”, the OB domain establishes contact with the backbone phosphates of the incoming DNA substrate (Akey et al., 2006; Unciuleac et al., 2019). While these interactions between the OB domain and ATP are not directly observed in either structure presented here, a structure of the LIG domain with ATP and MES suggests these interactions occur and that the interaction is dynamic (Unciuleac et al., 2019). Time-resolved x-ray crystallography may be a solution to acquire these intermediate conformations.

Capturing the active site in the pre-adenylylation state required mutating the catalytic lysine (K481) to methionine, to prevent formation of the LIG-AMP intermediate (Akey et al., 2006). A closer look at the pre-adenylylation state shows an ATP molecule, along with two magnesium ions in the active site (Figure 4D) (Akey et al., 2006). One magnesium ion is catalytic and is proposed to stabilize the transition state of the ATP by coordinating the ATP α-phosphate. This magnesium ion is also coordinated directly and indirectly through a water molecule by the conserved Glu613 (Akey et al., 2006). The second magnesium ion is thought to be non-catalytic, connecting with the β- and γ-phosphates of ATP. Amino acids surrounding the active site also stabilize the ATP molecule primarily through interactions with the α, β, and γ phosphates, and include amino acids His465, Arg501, Lys635, Lys637, and Arg629, while Arg486 forms a hydrogen bond with the ribose oxygen (Figure 4D) (Akey et al., 2006; Shuman and Glickman, 2007). It has been proposed that Lys635, Lys637 and the catalytic magnesium stabilize the adenylylation transition state, while the second metal ion works alongside three basic residues–Arg501, Arg629, and Lys635, to align the ATP γ-phosphate group for an in-line attack on ATP by the Lys481 nucleophile (Akey et al., 2006; Unciuleac et al., 2019). Mutational studies further confirmed the importance of these active site residues, where LigD with mutations Glu613Ala or Lys637Ala lacked nick-sealing activity (Akey et al., 2006), while LigD with mutations Arg501Ala or Arg629Ala had reduced nick-sealing activity compared to wild-type LigD (Unciuleac et al., 2019).

Once the adenylylation is complete and the LIG-AMP complex is formed, the active site residues still play a key role, but with some changes (Figure 4E) (Akey et al., 2006). A phosphoramidate bond is formed between the catalytic Lys481 and the phosphate group of AMP. This intermediate is coordinated in the active site through hydrogen bonds between a ribose sugar oxygen with Arg486 as in the pre-adenylylation state, while Arg501 shifts from the ATP γ-phosphate to the ribose sugar (Akey et al., 2006). A water-mediated interaction with Glu530 rounds out the hydrogen bonds to the ribose sugar. New hydrogen bonds also form in the active site between side-chains of Glu479 and Lys618, the backbone carbonyl of Gly480, and the backbone amide of Trp42 with the adenine base (Akey et al., 2006). These interactions highlight the substrate specificity of the active site for ATP.

From these crystal structures, some questions remain. Within the structure of the LIG-AMP intermediate, the electron density suggested the presence of a metal ion, although the question of whether one or both ions would fit in the active site is unclear. Additionally, LigD and the human homolog Ligase IV both preferentially ligate a nick containing a 3′ ribonucleotide (Zhu and Shuman, 2008; Pryor et al., 2018). While there are no structures of the LIG domain from LigD with DNA, Zhu and Shuman (2008) postulate that based on structures of Escherichia coli LigA, human LigI and Chlorella virus ligase bound to DNA nicks, the 3′-hydroxyl group is forced into an A-like conformation, similar to RNA. It is possible that bacterial LigD cannot restructure DNA into an A-form, thus requiring the presence of a 3′-ribonucleotide for optimal ligation (Zhu and Shuman, 2008). This theory is further supported by biochemical studies showing that the LIG domain alone was 12-fold faster for sealing a nick with a ribonucleotide compared to a nick with deoxyribonucleotides (Unciuleac et al., 2019). Future crystal structures of the LIG domain in complex with ribonucleotide-containing substrates will help solve this puzzle.

The polymerase domain of LigD belongs to the archaeo-eukaryotic primase (AEP) superfamily (Iyer, 2005). The basic function of the POL domain of LigD is that of a nucleotidyltransferase, where a phosphodiester bond is created between an incoming nucleotide and the 3′-hydroxyl end of a DNA primer strand. At least 2 divalent metal ions are required, coordinated by conserved aspartate residues within the polymerase active site. One divalent metal ion primes the 3′-hydroxyl group of the primer strand for nucleophilic attack on the γ-phosphate of the incoming nucleotide, while the second divalent metal ion aids in the removal of the pyrophosphate, thus creating the phosphodiester bond between the primer strand, and incoming nucleotide (Beese and Steitz, 1991; Steitz, 1993; Steitz, 1999). The LigD POL domain is unique compared to other polymerases, containing a repertoire of nucleotidyltransferase functions not found in other members of the AEP superfamily, which typically function solely as primases during DNA replication (Pitcher et al., 2005; Guilliam et al., 2015). More specifically, the LigD POL domain is capable of DNA or RNA gap-filling, RNA primase, and terminal transferase activities (Weller et al., 2002; Della et al., 2004; Gong et al., 2004; Gong et al., 2005; Pitcher et al., 2005; Yakovleva and Shuman, 2006).

As a member of the AEP superfamily, the LigD POL domain shares a common catalytic core, consisting of a N-terminal (αβ)₂ unit that is packed onto a derived C-terminal RNA recognition motif (RRM) (Figure 5A) (Guilliam et al., 2015). Within this catalytic core lies three highly conserved motifs–an hhhDhD motif (where “h” is a hydrophobic residue, motif I), an sxH motif (where “s” is a small residue and “x” can be anything, motif II), and an hD/E motif (motif III) (Iyer, 2005; Guilliam et al., 2015). Motifs I and III are responsible for coordinating divalent metal ions during enzyme catalysis, whereas motif II binds the incoming nucleotide (Guilliam et al., 2015). Although the POL domain has considerable homology with replicative primases, its specialized role in DSB repair is attributed to its possession of unique structural elements in the regions following motif I, and in between motifs II and III (Koonin et al., 2000). These elements include a phosphate-binding pocket, in addition to a pair of distinct surface loops that aid with the synapsis of DNA breaks during NHEJ repair (Iyer, 2005).

Crystal structures of the POL domain from both P. aeruginosa and M. tuberculosis have been solved in the apo form (PDB 2FAO, 2IRU), and in the presence of nucleotides and divalent metal ions (PDB 2FAR, 2FAQ, 2IRY, and 2IRX) (Zhu et al., 2006; Pitcher et al., 2007b). The LigD POL domain apo structures from both species are well aligned (RMSD = 1.5 Å) (Figure 5A) and additional structures with DNA substrates have been solved from M. tuberculosis, therefore we will focus our attention on the M. tuberculosis structures.

The catalytic site of the POL domain is found in a cleft on the surface of the protein, surrounded by a positively charged region (Figure 5B), that functions both to align the incoming DNA substrate and mediate interactions between homodimers of the POL domain as will be discussed below (Brissett et al., 2007; Pitcher et al., 2007b; Brissett et al., 2013). In the presence of dGTP, the POL domain crystallized with a single divalent metal ion, manganese. Octahedral coordination of the manganese occurs with oxygen atoms of the conserved Asp137(motif I) and Asp139 (motif I) sidechains, oxygen atoms from the β- and γ-phosphates of dGTP and neighboring water molecules (Figure 5C) (Zhu et al., 2006; Pitcher et al., 2007b). The dGTP is further stabilized by a conserved triad of residues, Ser172 (motif II), His178 (motif II), and Arg244 with hydrogen bonds between the sidechains of the residues and oxygen atoms of the incoming nucleotide’s phosphate groups. A conserved Phe64 also base stacks with the guanosine ring (Pitcher et al., 2007b). Additional water molecules are involved in the hydrogen-bonding network that stabilizes the active site but have been left out of Figure 5C for clarity. Details of this hydrogen-bonding network are available in previous publications (Pitcher et al., 2007b).

Missing from this structure was the presence of the second metal ion required for catalysis. However, crystallization of the LigD POL domain with UTP, manganese and DNA captured both metal ions in the active site (PDB 3PKY) (Brissett et al., 2011). The second metal ion forms a tetrahedral coordination with the same conserved aspartate residues, but also includes the side chain of an additional aspartate (Asp227, motif III) (Figure 6A). This interaction would prime the 3′-hydroxyl end of the DNA template strand, which is absent in this structure. The coordination of the manganese ion transforms the active site into a pre-catalytic complex, allowing UTP to base-pair with the adenine base of the templating DNA strand (Brissett et al., 2011). The conserved Phe64 base stacks with the templating DNA strand, rather than the incoming nucleotide as in Figure 5C (Figure 6B). New interactions in the active site further stabilize UTP, with hydrogen bonds forming between the 2′-hydroxyl group of the ribose, Thr236, and His111. An overlay of the active sites shows the changes of the amino acids in the active site when bound to dGTP vs the UTP:manganese:DNA complex. Of interest is that while a second divalent ion is observed in the tripartite complex, the triphosphate tail of dGTP likely occluded binding of the second manganese ion in the structure of the POL domain with dGTP (Brissett et al., 2011).

The ability of the LigD POL domain to insert ribonucleotides preferentially is of interest (Della et al., 2004; Pitcher et al., 2007a; Zhu and Shuman, 2010). Biochemical studies in P. aeruginosa show that the His111 homolog, when mutated in P. aeruginosa LigD, has reduced preference for ribonucleotides compared to wild-type LigD (Sánchez-Salvador and De Vega, 2020). In both tripartite complex structures from M. tuberculosis and P. aeruginosa, the conserved histidine hydrogen bonds to the 2′-hydroxyl of the ribose, dictating selectivity for ribonucleotides in the active site (Sánchez-Salvador and De Vega, 2020).The POL domain structure also plays a role in maintaining fidelity. P. aeruginosa LigD showed decreased polymerization efficiency and increased nucleotide misincorporation with a mutation of Lys606 to alanine (Sánchez-Salvador and De Vega, 2020). This lysine is conserved and in M. tuberculosis, the homologous Lys66 sandwiches the base of either the incoming nucleotide or templating DNA strand with Phe64 (Figure 6B) (Pitcher et al., 2007b; Brissett et al., 2011). While the exact mechanism of fidelity is unclear, it is hypothesized that the lysine helps select the correct nucleotide for insertion, when a DNA template is present (Sánchez-Salvador and De Vega, 2020).

The LigD POL domain needs to be capable of handling complex DNA DSBs. While a gapped DNA substrate contains a connected template and primer strand, in the case of a DSB, the physical connection is severed, creating discontinuous template and primer strands. The mechanism by which LigD handles this challenge is to form a synaptic complex using two LigD proteins, as illustrated in crystal structures of the LigD POL domain in complex with DNA DSBs (Figure 7) (Brissett et al., 2007, 2013). We aligned the structure of the POL:UTP:DNA:Mn²⁺ with each synaptic complex (RMSD = 0.37 Å for Figure 7A, RMS = 0.34 Å for Figure 7B) to highlight the active site and relative location of the incoming nucleotide. How these POL domains interact with one another is dependent on whether the DNA ends are complementary or non-complementary. A catalytically competent synaptic arrangement is formed when bound to a DNA substrate with self-complementary ends (Figure 7A) (Brissett et al., 2013). Here, the POL domains are rotated 180^o around the y-axis with respect to one another, where the template strand from one POL domain becomes the incoming primer strand for the adjacent POL domain. Microhomologies between the 3′ ends of each incoming DNA strand helps stabilize the synapsis and provides the necessary primer strand that allows for an in trans polymerization mechanism. In essence, this configuration mimics gapped DNA substrates (Brissett et al., 2013). Critical to the formation of this complex is the phosphate binding pocket made up of Lys16, Lys26, and Asn13 on LigD, to bind a downstream 5′-phosphate group, where only one 5′-phosphate group is required for synapsis (Brissett et al., 2013). Conserved loops 1 and 2 of the POL domain also stabilize the synaptic complex (Brissett et al., 2007; Brissett et al., 2013), while loop 1 additionally interacts with the template strand to re-orient it to become a primer strand, while loop 2 guides the 3′ DNA end into the polymerase active site (Brissett et al., 2013). In contrast, when the POL domain is bound to a non-complementary DNA substrate, a synapse is also created, but one that is catalytically incompetent for polymerization extension (Figure 7B). In this structure, the POL domains are rotated 180^o around the x-axis with respect to each other, with the 3′-DNA ends of opposing DNA substrates forming a synaptic complex (Brissett et al., 2007). The 5′-phosphate of the downstream DNA strand is still bound by the phosphate-biding pocket, and loop 1 is critical to forming the synaptic complex and re-orienting the incoming 3′-end, with loop 2 also contacting the DNA (Brissett et al., 2007). However, because the DNA ends are not complementary, the DNA is distorted, with only small microhomologies being formed. These distortions result in the 3′ ends failing to reach the catalytic site for extension, instead forming hairpin structures. It is hypothesized that the space between the synapsed POL domains could accommodate the PE domain and allow additional processing of the DNA ends prior to polymerization (Brissett et al., 2007).

When first discovered, the PE domain lacked significant homology to any known family of bacterial nucleases, and thus was classified as a new phosphoesterase family (Zhu and Shuman, 2005b). Since then, this phosphoesterase domain has also been found in proteomes of both archaea and eukaryotes (Nair et al., 2010). The PE domain is multifunctional–reflective of its role in a multipurpose enzyme. It is a phosphoesterase, converting 3′-phosphate ends to the necessary 3′-hydroxyl for ligation, while also being a ribonuclease, resecting lengths of 3′-ribonucleotides introduced by the POL domain, and leaving the preferred single 3′-ribonucleotide for ligation (Zhu and Shuman, 2005b, 2008). The mechanism by which the PE domain carries out 3′ end-healing activities occurs in two manganese-dependent steps (Zhu and Shuman, 2005b). First, removal of a 3′-terminal nucleoside on the primer strand leaves behind a ribonucleoside with a 3′-phosphate group (phosphodiesterase activity). Then, via hydrolysis, the 3′-phosphate group is converted to a 3′-hydroxyl group with the release of an inorganic phosphate (phosphomonoesterase activity) (Zhu and Shuman, 2005b; Zhu and Shuman, 2006; Zhu et al., 2005; Nair et al., 2010). These activities create a comprehensive 3′ end-processing mechanism in the PE domain necessary for DSB repair and is found in P. aeruginosa, A. tumefaciens, and mycobacterial LigD (Zhu and Shuman, 2005b; Zhu and Shuman, 2006; Zhu and Shuman, 2007).

Atomic structures of the P. aeruginosa LigD PE domain have been solved by both x-ray crystallography and nuclear magnetic resonance (PDB 3N9B, 2LJ6) (Nair et al., 2010; Natarajan et al., 2012). Additional structures of archaeal PE domains from Candidatus Korarchaeum cryptofilum and Methanosarcina barkeri are more compact versions of P. aeruginosa PE domain, although the catalytic core is conserved (Nair et al., 2010). The Methanocella paludicola PE domain (PDB 5DMP) (Bartlett et al., 2016), though, aligns well with the P. aeruginosa PE domain (RMSD = 0.45 Å), therefore we will focus the remainder of our discussion on the P. aeruginosa PE domain. The PE domain forms an eight-stranded beta barrel, that is bounded by two alpha helices and a 3₁₀ helix (Figure 8A) (Nair et al., 2010). The beta barrel structure is maintained through a cluster of conserved hydrophobic residue pairs, while the hydrophilic active site is situated on the exterior of the barrel, in a crescent-shaped groove (Figure 8C) (Nair et al., 2010). Alongside the active site runs a stretch of positive charge (Figure 8B), which could interact with the negatively charged backbone of DNA, guiding the 3′-DNA end into the active site for processing.

Within the structure of the active site are the required manganese ion and a sulfate anion that can be considered a mimetic for the scissile phosphate to be cleaved in the DNA backbone (Nair et al., 2010). The manganese is stabilized by an octahedral coordination complex, mediated by a catalytic triad of conserved residues–His42, His48, and Asp50–that when expressed as alanine point mutations in P. aeruginosa LigD, are catalytically dead (Zhu and Shuman, 2006). Two neighboring water molecules and the oxygen from the sulfate anion complete the coordination of the manganese ion (Nair et al., 2010). The sulfate anion is proposed to stabilize the transition state and is coordinated by another highly conserved group of residues, Arg52, His84, and Tyr88. Like the catalytic triad surrounding the manganese, point mutations of His84, and Tyr88 in P.aeruginosa LigD are also catalytically inactive, highlighting the necessity of these residues in the phosphoesterase active site (Zhu and Shuman, 2006). Also within the active site are two more conserved residues, Gln40, and Ser 61. While these residues do not appear necessary for the catalytic mechanism, they may be key to forming the structure of the active site, along with Arg52, as LigD proteins with a Gln40Ala mutation had severely reduced phosphodiesterase activity compared to wild-type LigD (Zhu and Shuman, 2006; Nair et al., 2010).

Biochemical studies have also identified that Arg14, Arg15, Glu21, and Glu82 are necessary for 3′-phosphatase activity, but not removal of the 3′-ribonucleoside (Zhu and Shuman, 2006). The N-terminus of the PE domain is disordered in structures from both x-ray crystallography and nuclear magnetic resonance (Nair et al., 2010; Natarajan et al., 2012), precluding structural information about Arg14, Arg15, and Glu21. However, Glu82 resides on the outer edge of the active site. While in the current structure of the PE domain, the side chain of Glu82 is pointed away from the active site, it is likely that in the presence of the DNA substrate, Glu82 may become more involved, given its relevance in the biochemical activity of the PE domain.

Absent in the crystal structure of the PE domain is a loop outside the active site (Figure 8A, dashed line), which is seen in crystal structures of the archaeal PE domain from Methanocella paludicola (Bartlett et al., 2016). Based on a lack of density for this loop in the P. aeruginosa PE domain structure (Nair et al., 2010), coupled with evidence from nuclear magnetic resonance studies showing that this loop moves away from the active site when DNA is present (Natarajan et al., 2012), it is likely this loop is flexible. In the M. paludicola PE domain, the loop appears to cover the active site, which contains a magnesium and vanadate ion (Bartlett et al., 2016). A similar arrangement is likely for the structure of the PE domain in P. aeruginosa, given the current trajectory of the ends of the loop that are visible, and the presence of similar substrates in the active site. Future structures of this domain with DNA may well show that this loop acts as a capping mechanism, allowing access to the active site only in the presence of the correct DNA substrate.

Atomic structures of the LIG, POL, and PE domain from bacteria and archaea, combined with biochemical studies, continue to be instrumental in highlighting the structural foundation for enzymatic activity in LigD. However, the arrangement of these domains in three-dimensional space in wild-type LigD is unknown, along with how the structural arrangement affects processing of the DNA DSB. This idea is especially interesting, given that the primary structure of the domains can vary between bacterial species (Figure 3). Until an experimental structure is obtained, either by cryo-electron microscopy or x-ray crystallography, recent advances in in silico structure predictions using the AlphaFold algorithm (Jumper et al., 2021) through the ColabFold notebook (Mirdita et al., 2021), provide an opportunity to explore possible models of wild-type LigD.

We generated predictions of LigD from P. aeruginosa, and B. subtilis using the ColabFold notebook, and downloaded the prediction of M. tuberculosis LigD from the AlphaFold Protein Structure Database (Figure 9) (Jumper et al., 2021; Mirdita et al., 2021). Interestingly, both P. aeruginosa, and B. subtilis LigD have compact conformations, with the enzymatic domains arranging into an almost globular structure. M. tuberculosis though, is slightly more elongated, forming a curved, horseshoe like structure (Figures 9A,C,G). Are these compact folds functionally active in the LigD models? The LigD LIG domain contains an NTase and OB domain, where the OB domain is flexible, and existing in both open and closed conformations as described earlier. We overlaid the structure of the M. tuberculosis LIG domain in the open conformation, using the NTase domain for alignment (PDB 1VS0) (Akey et al., 2006) with each LigD model. We found that for M. tuberculosis and B. subtilis LigD models, the OB fold is in between the open and closed conformation, however for the P. aeruginosa LigD model, the OB fold is closed, capping the active site. Interestingly, if the OB fold were to open, it would collide with the POL domain in the current model (Figure 9D). Therefore, some flexibility must be inherent in these structures, at the least in P. aeruginosa LigD to accommodate accessibility of the active site. Flexibility is also essential if the structure of the POL domain bound to a synaptic DNA substrate (PDB 4MKY) (Brissett et al., 2013) is considered as well. Homodimerization of LigD, via the POL domain bound to a DNA synapse would not cause any conflicts in M. tuberculosis or B. subtilis LigD, however in the present model of P. aeruginosa, the POL domain would collide with the LIG domain (Figure 9E). Long loops predicted between the POL, LIG, and PE domains are likely to be flexible, based on the lower confidence scores in positioning these loops by ColabFold and would permit alternate conformations of P. aeruginosa LigD. Small-angle x-ray scattering studies may be able to answer these questions about the movement of LigD in solution.

We also aligned structures of the LIG, POL, and PE domains containing substrates in their active site, with the LigD models to better visualize how repair may occur (Figures 9B,F,H). For the LIG domain, we used the substrates of ATP and magnesium (PDB 6NHZ) (Unciuleac et al., 2019), creating an alignment through the NTase fold, while for the POL domain, we used the complex containing DNA, UTP, and manganese (PDB 3PKY) (Brissett et al., 2011). For the PE domain, we used the substrates of manganese and the sulfate ion (PDB 3N9B) (Nair et al., 2010) in the alignment. As shown in the electrostatic surface models (Figures 9B,F,H), the active sites of each domain are on the outer surface and accessible to the solvent, with the exception of the trapped LIG active site in P. aeruginosa, and in the absence of a synaptic complex in the P. aeruginosa POL domain. The electrostatic surface shows trails of positive regions leading to the active sites and may form a guiding path for the DNA ends in need of polymerase extension, ribonucleoside removal or ligation of the DSB. This path would allow a DNA DSB to move between the different enzymatic domains, depending on the processing requirements at the DSB, with active site structures allowing or denying access to the DSB, depending on whether the DNA substrate is a fit. Alternately, LigD may move around the DSB, capturing and releasing the ends from the active site of each domain, carrying out repair. Ku may also have a role in coordinating LigD-DNA binding. M. tuberculosis Ku binds the POL domain of LigD and weakly to the LIG domain, which could change the conformation of LigD to better accommodate DNA ends (Della et al., 2004; Pitcher et al., 2005; Wright et al., 2010). Answers to these questions and more are to be found in future experimental structures of LigD, bound to Ku and/or in the midst of repairing a DNA DSB.