Eukaryotic DNA Ligases: Structural and Functional Insights

DNA Ligase I Family

The CDC9 gene, which encodes the Saccharomyces cerevisiae homolog of human LIG1, was first identified in the screen for cell division cycle mutants (36). The CDC9 gene is essential for viability, and temperature-sensitive alleles are hypersensitive to DNA damage, implicating the Cdc9 DNA ligase in DNA replication and repair (37, 38). Interestingly, an alternative translation initiation mechanism generates two forms of Cdc9 DNA ligase (Figure 4) that are directed to either the nuclear or the mitochondrial compartment and are essential for genome maintenance in these organelles (39). The conditional lethal phenotype of a cdc9 mutant strain was utilized to identify cDNAs, encoding human DNA ligase I, that permitted growth at the nonpermissive temperature (4). Unlike the yeast CDC9 gene, the human LIG1 gene does not encode a polypeptide that is targeted to mitochondria. In vertebrates, the LIG3 gene encodes the mitochondrial DNA ligase (see below).

As shown schematically in Figure 4, the polypeptides encoded by the human LIG1 and the yeast CDC9 genes contain a highly related C-terminal catalytic core. The adjacent N-terminal regions of these enzymes share very little overall amino acid sequence homology and are not required for catalytic activity (4, 26). This N-terminal region of human DNA ligase I was not present in the catalytic fragment that was crystallized with nicked DNA (26), and it is not required for complementation of yeast cdc9 mutant strains by ligase I. Notably, this region contains sequences that are required for nuclear localization and targeting to replication factories (40–42), and it is involved in protein-protein interactions described below.

The N-terminal region of DNA ligase I is also subject to posttranslational modification. Serine residues at positions 51, 76, and 91 (Figure 4) are phosphorylated by cyclin-dependent kinase (43), and Ser66 is phosphorylated by casein kinase II (44). From low levels in the G1 phase of the cell cycle, phosphorylation increases during S phase, reaching a peak at G2/M (44). Phosphorylation of Ser91 occurs at the G1/S transition and is necessary for subsequent phosphorylation at Ser76 in the hyperphosphorylated G2/M form of DNA ligase I (43). Interestingly, the targeting of DNA ligase I to replication factories was markedly reduced when serine residues 51, 66, 76 and 91 were replaced with aspartic acid (43). Because DNA ligase I is dephosphorylated in response to both DNA damage and replication blockage (45), it appears that the participation of this enzyme in different DNA transactions is regulated by posttranslational modifications that presumably modulate interactions with different protein partners. Recent analyses of the phosphoproteome have identified additional phosphorylation sites in the N-terminal region of DNA ligase I and also in the C-terminal region beyond the catalytic core (46, 47), indicating a complex regulation by post-translational modifications.

DNA ligase I-deficient cell lines have been established from the only known case of a human with mutated LIG1 alleles (48, 49). The clinical symptoms of this individual are described in the next section. One of the mutant LIG1 alleles was inherited, whereas it is not known how the second mutant allele was acquired (48, 49). In the inherited maternal allele, the mutation results in the replacement of a conserved residue, Arg771, with a tryptophan residue within the OB domain. Although this mutant polypeptide is defective in the second step of the ligation reaction, it retains partial activity—about 10% compared with wild-type DNA ligase I (48, 50). In the other mutant allele, the mutation results in the replacement of glutamic acid 566 with a lysine residue. The nonconservative replacement of a conserved residue two amino acids away from the key active-site lysine residue almost certainly results in enzyme inactivation. Notably, although the presumed null allele has been lost from the SV40-immortalized derivative (46BR.1G1) of the primary fibroblast cell line (46BR) established from the DNA ligase I-deficient individual, the primary and immortalized fibroblast cell lines have similar phenotypes (48, 51). The 46BR.1G1 cell line has a modest reduction (about two-fold) in the amount of DNA ligase I protein compared with other SV40-immortalized fibroblasts, but a more dramatic reduction in DNA ligase I activity (10- to 20-fold) (48). In accord with its expected role in DNA replication, the DNA ligase I-deficient cells exhibit a marked defect in Okazaki fragment joining, but unexpectedly, this does not have an obvious effect on cell proliferation (52, 53). It is notable that 46BR.1G1 cells are hypersensitive to 3-amino benzamide (51), an inhibitor of poly(ADP-ribose) polymerase, which participates, together with DNA ligase IIIα, in the repair of DNA single-strand breaks (54).It appears likely that the poly(ADP-ribose) polymerase 1 (PARP1)-dependent pathway acts on single-strand breaks remaining in the lagging strand because of the DNA ligase I deficiency. The 46BR.1G1 cells are hypersensitive to simple DNA-alkylating agents because of a defect in the long-patch subpathway of base excision repair (55). In addition, 46BR cells exhibit modest sensitivity to other DNA-damaging agents, including UV-light and ionizing radiation (48, 49). An interesting, and as yet unexplained, feature of the 46BR.1G1 cells is that they are hypomutable following DNA damage (52).

Because CDC9 is an essential gene in S. cerevisiae, it was assumed that inactivation of the mammalian LIG1 gene would result in cell lethality. This notion was supported by the observation that loss of both LIG1 alleles in mouse embryonic stem cells was only observed in cells ectopically expressing full-length DNA ligase I cDNA (56). Thus, it was surprising that mouse embryos harboring homozygous deletions of the 3′ end of the LIG1 gene developed normally until midgestation (57). It is possible that the embryos were not totally devoid of DNA ligase I activity because the gene-targeting strategy did not disrupt the LIG1 gene by deleting the first exon. However, a similar result was obtained using a different targeting strategy (although again not targeting the first exon) (58). Mouse embryonic fibroblast (MEF) cell lines established from both of these animal models proliferate normally but have a defect in Okazaki fragment joining and increased genomic instability (57, 58). Thus, there is strong evidence that mammalian DNA ligase I is not absolutely required for DNA replication and for the early stages of embryo development. Recently, the Melton laboratory established a mouse knockin model that expresses a mutant lig1 allele corresponding to the 46BR.1G1 mutant allele in which Arg771 is replaced by Trp (59). The phenotype of these animals is described in the next section. As expected, MEFs established from these animals are defective in Okazaki fragment joining (59). Unlike the 46BR.1G1-immortalized human fibroblast cell line, neither these MEFs nor the MEFs established from the lig1 null mice are sensitive to DNA-damaging agents, including DNA-alkylating agents (57–59). It is possible that differences in the relative contribution of the short- and long-patch sub-pathways of base excision repair (BER) underlie the different DNA damage sensitivities of human and mouse cell lines expressing the same mutant version of DNA ligase I.

DNA Ligase III Family

As mentioned previously, the LIG3 gene appears to be restricted to vertebrates. Thus, genetically tractable lower eukaryotes such as S. cerevisiae cannot be used to delineate the cellular functions of the LIG3 gene. Moreover, conclusions regarding the cellular functions of the CDC9 and DNL4 genes, the yeast homologs of mammalian LIG1 and LIG4, cannot be extrapolated to mammalian cells because of their larger repertoire of DNA ligases. The LIG3 gene of higher eukaryotes generates three distinct DNA ligase polyeptides (Figure 5). Nuclear and mitochondrial versions of DNA ligase IIIα are ubiquitously synthesized from DNA ligase IIIα mRNA by an alternative translation initiation mechanism (60, 61). In addition, a germ cell-specific alternative splicing mechanism, in which the terminal 3′-coding exon in DNA ligase IIIα mRNA is replaced by a different exon, generates DNA ligase IIIβ mRNA (61, 62).

The polypeptides encoded by the mammalian LIG3 gene contain a catalytic region that is most closely related to that of the DNA ligases encoded by the pox viruses (5, 6). This level of homology (about 50% sequence identity) extends over the entire length of the vaccinia DNA ligase (Figure 5). As mentioned above, a unique feature of the DNA ligases encoded by the LIG3 gene is an N-terminal zinc finger, which is related to the tandemly arrayed zinc fingers at the N terminus of PARP1 (6). Although the DNA ligase III zinc finger is not required for catalytic activity, efficient DNA joining by DNA ligase III at physiological salt concentrations is dependent upon this DNA-binding region (31).

As mentioned previously, the DNA ligase IIIα and β polypeptides have different C termini (61, 62). The 77 residues that are unique to DNA ligase IIIα constitute a breast and ovarian cancer susceptibility protein 1 C-terminal (BRCT) motif, a conserved sequence found in other proteins involved in cell cycle regulation and DNA repair that was originally identified in the C-terminal region of the breast cancer susceptibility protein BRCA1 (63). In DNA ligase IIIβ, these residues are replaced by a short sequence of 17–18 residues, which acts as a nuclear localization signal (Figure 5). The BRCT motif is often involved in protein-protein interactions and, in the case of the nuclear form of DNA ligase IIIα, it interacts with the DNA repair protein XRCC1, resulting in the formation of a DNA ligase IIIα-XRCC1 complex (62, 64–66). Because XRCC1 is required to stabilize DNA ligase IIIα, XRCC1-deficient cell lines, such as the Chinese hamster ovary cell lines EM9 and EMC11, have reduced levels of DNA ligase IIIα and so are functionally DNA ligase IIIα deficient (67). The phenotype of XRCC1-deficient cell lines includes pronounced hypersensitivity to DNA-alkylating agents, modest hypersensitivity to ionizing radiation, and spontaneously elevated levels of sister chromatid exchange (68). However, it cannot be assumed that the phenotype of DNA ligase III mutant cells will recapitulate that of xrcc1 mutant cells because XRCC1 also functions independently of DNA ligase IIIα in nuclear DNA repair (69) and because DNA ligase IIIα functions independently of XRCC1 in mitochondrial DNA metabolism (70).

Deletion of either the XRCC1 or the LIG3 gene results in early embryonic lethality in the mouse, with the lig3 null embryos surviving 1–2 days longer than the xrcc1 null embryos (71, 72). Attempts to establish MEF cell lines from these embryos have failed. However, the absence of p53 extended embryonic development of xrcc1 null embryos, facilitating the establishment of mouse p53^−/− xrcc1^−/− embryonic fibroblasts (71) and suggesting the absence of XRCC1 and presumably DNA ligase III, resulted in the persistence of DNA damage that activated p53-dependent pathways for cell cycle checkpoints and/or apoptosis. On the basis of these results, it is evident that XRCC1 is not required for cell viability. Because mitochondrial DNA ligase IIIα functions independently of XRCC1 (70), it is possible that total inactivation of the LIG3 gene will result in cell lethality owing to defects in mitochondrial DNA metabolism. Indeed, using RNAi to reduce the cellular level of DNA ligase IIIα, Campbell and colleagues (73) showed that mitochondrial DNA ligase IIIα is critical for the function of these organelles.

In accord with the hypersensitivity of xrcc1^−/− cells to DNA-alkylating agents, there is compelling evidence from biochemical and cell biology studies linking the nuclear DNA ligase IIIα-XRCC1 complex with a subpathway of BER (short patch) that is characterized by a single-nucleotide repair tract (74) and the repair of single-strand breaks (75, 76). More recently, it has been shown that the nuclear DNA ligase IIIα-XRCC1 complex is a key component of a nucleotide excision repair (NER) subpathway that operates in dividing cells and is particularly important for the repair of UV lesions in nondividing cells (77). Thus, in contrast to S. cerevisiae where Cdc9 is the only ligase active in excision repair, both DNA ligase I and DNA ligase IIIα participate in mammalian excision repair. There is also evidence that DNA ligase IIIα is a component of a backup pathway for the major DNA ligase IV-dependent nonhomologous end-joining pathway (78). Finally, it is possible that the presence of DNA ligase IIIα may compensate for the absence of DNA ligase I in lig1 null cells, allowing these cells to replicate and survive. The elucidation of the cellular and organismal roles of the nuclear and mitochondrial DNA ligase IIIα and the germ cell-specific DNA ligase IIIβ will require more sophisticated gene-targeting strategies or specific inhibitors of these enzymes.

DNA ligase IIIα is phosphorylated on Ser123, located between the N-terminal zinc finger and the catalytic region of DNA ligase III (79). This residue is progressively phosphorylated by the cyclin-dependent kinase (Cdk2) as cells progress through S phase, and it is dephosphorylated in response to DNA damage in a reaction dependent upon ATM, the ataxia-telangiectasia-mutated protein (79). The regulation of the steady-state level of phosphorylation by ATM appears to involve both inhibition of Cdk2 activity and activation of a phosphatase, most likely protein phosphatase PP1 (79). Given the marked sensitivity of cell lines deficient in DNA ligase IIIα to DNA-alkylating agents (68), it is surprising that oxidative stress, rather than DNA alkylation damage, is the most effective inducer of DNA ligase IIIα dephosphorylation (79). Analysis of the phosphoproteome has identified additional sites within the N-terminal region of DNA ligase IIIα (47). As described below, the region of DNA ligase III containing the Ser123 phosphorylation site mediates interactions with different protein partners, raising the possibility that these interactions are regulated by posttranslational modification.

DNA Ligase IV Family

In contrast to the other human DNA ligases, the DNA ligase encoded by the LIG4 gene was initially identified by a genomics approach in which a human expressed sequence tags database was searched with a sequence motif that is conserved within the catalytic domain of eukaryotic DNA ligases (6). As shown schematically in Figure 6, the polypeptides encoded by the human LIG4 and the yeast DNL4 genes have an N-terminal catalytic domain with a C-terminal extension that contains tandemly arrayed BRCT motifs (6, 80–82). Because a large fraction of DNA ligase IV molecules purified from cultured cells are adenylated, this enzyme is not robustly labeled in assays measuring enzyme-AMP formation using [α³²P]ATP (83). This characteristic, combined with the similar size of the polypeptides encoded by the LIG3 and LIG4 genes, probably explains the failure to detect DNA ligase IV activity prior to its cloning. Another noteworthy feature of DNA ligase IV that distinguishes it from DNA ligases I and III is its ability to join DNA ends that are noncomplementary because of mismatches and/or short gaps (84).

DNA ligase IV interacts with human XRCC4 (85, 86), a protein involved in the repair of DNA double-strand breaks by nonhomologous end joining (NHEJ) and the completion of V(D)J recombination (87). This complex provided the initial insights into the cellular function of DNA ligase IV. Similar to the nuclear form of DNA ligase IIIα (67), DNA ligase IV is unstable in the absence of XRCC4, and so xrcc4 cell lines are functionally deficient in DNA ligase IV activity (88). The stability of DNA ligase IV is also regulated by phosphorylation at Ser650. Notably, DNA ligase IV is phosphorylated by the DNA-dependent protein kinase, DNA-PK, a key NHEJ factor that negatively regulates DNA ligase IV stability (89). As expected, targeted inactivation of LIG4 in mouse and human cells dramatically reduced V(D)J recombination and resulted in hypersensitivity to ionizing radiation because of the defect in the repair of double-strand breaks by NHEJ (90, 91). In addition, MEFs established from lig4 null embryos have a growth defect and reduced life span owing to premature senescence (90). Inactivation of the mouse LIG4 gene results in late embryonic lethality (90, 92). Prior to their demise, the developing lig4 null embryos exhibit defective lymphopoiesis, which presumably reflects their inability to carry out V(D)J recombination, but these embryos appear to die because of massive apoptosis in the central nervous system (90, 92, 93). Notably, the excessive neuronal apoptosis and embryonic lethality, caused by lig4 deficiency but not by the lymphocyte development defects, were rescued by concomitant loss of either p53 or ATM function (35, 93, 94). Thus, it appears that DNA ligase IV is involved in the repair of endogenous DNA damage that if unrepaired elicits apoptosis in neuronal cells via a pathway mediated by p53 and ATM. At the cellular level, a p53 deficiency but not ATM deficiency corrects the proliferation defects in lig4 MEFs (93, 94). The clinical symptoms of the several human individuals with mutant lig4 alleles (95, 96) are described in the section below. As expected, cell lines established from these individuals are highly radiosensitive (95, 96).

In S. cerevisiae, inactivation of the DNL4 gene does not cause significant hypersensensitivity to ionizing radiation unless homology-dependent pathways of DNA double-strand break repair are also inactivated (80–82). This reflects differences in the relative contributions of NHEJ and homology-dependent repair pathways between mammals and S. cerevisiae. In fact, yeast dnl4 strains are more resistant to agents that cause DNA double-strand breaks because Dnl4 not only catalyzes the last step of NHEJ but also suppresses homologous recombination (97).

DNA Ligase I Deficiency

The only known case of a human DNA ligase I deficiency caused by an inherited mutant lig1 allele was diagnosed in a patient with recurrent infections. This symptom resulted from unexplained immunodeficiencies, including hypogammaglobulinemia and poor proliferative response of lymphocytes to mitogens in vitro (48, 49). The female patient also exhibited sensitivity to sunlight, growth retardation, and delayed development. Hepatosplenomegaly was detected at age 17 with further analysis revealing lymphocyte infiltration of portal tracts, indicative of lymphoma. This individual died at age 19 from pneumonia. Although DNA ligase I-deficient cell lines established from this patient exhibit defects in DNA replication and repair (48, 50, 52, 55, 98), there is no defect in V(D)J recombination to explain the immunodeficiency (99, 100). It appears likely that the immunodeficiency was caused by a defect in the proliferation of B and T cells in response to stimuli. With only a single case, it is not possible to fully characterize the clinical symptoms caused by DNA ligase I deficiency in humans. Notably, a mouse model has been established reiterating the Arg771 to Trp change found in the DNA ligase I-deficient human (59). As in the human, this resulted in delayed growth. In young mutant mice, hematopoietic defects were observed, including enlargement of the spleen (59). The mutant mice did not exhibit signs of a compromised immune system, although this may result from housing the animals in a clean environment. Increased levels of spontaneous genomic instability were evident in spleen cells from the mutant mice, and there was also a marked increase in the incidence of spontaneous tumors in the mutant mice (59). Thus, it appears that the DNA replication defect caused by DNA ligase I deficiency in mammals results in delayed growth and increased cancer incidence because of genomic instability. This proliferation defect also appears to underlie abnormalities in hematopoiesis and lymphopoiesis (49, 59).

DNA Ligase IV Deficiency

Five individuals with DNA ligase IV deficiency have been identified (95, 96, 101). These individuals exhibited a range of clinical symptoms (95, 96, 101) that appear to reflect differences in the residual DNA ligase IV activity encoded by the hypomorphic lig4 alleles (102). Notably, reduced DNA ligase IV activity has a more dramatic effect on the repair of double-strand breaks by NHEJ than V(D)J recombination (95, 96). The first case of DNA ligase IV deficiency was detected because of an extreme reaction to radiation treatment for leukemia (96). In contrast to the first DNA ligase IV-deficient individual, who apart from the leukemia was clinically normal, the other four individuals had more severe and diverse symptoms that included distinctive facial features, microcephaly, skin abnormalities, growth and developmental deficiencies, and immunodeficiency (95, 101). In the DNA ligase IV-deficient syndrome, the immunodeficiency is caused by defective V(D)J recombination. There is considerable overlap between the symptoms associated with DNA ligase IV deficiency and those of the DNA damage response disorder, Nijmegen breakage syndrome (95, 103, 104). Although these syndromes are characterized by radiation sensitivity at the cellular level, the radiosensitivity in nbs1 mutant cells is primarily a consequence of a checkpoint defect, whereas it is caused by a defect in the repair of DNA double-strand breaks by NHEJ in lig4 mutant cells (95, 96). At the present time, there are too few cases to determine whether DNA ligase IV deficiency causes cancer predisposition in humans. In the mouse, however, there is evidence that DNA ligase IV deficiency increases cancer formation in nonlymphoid tissues. Specifically, lig4 heterozygosity in the tumor-prone ink4a/arf mouse strain promotes the development of soft tissue sarcomas (105).

DNA Ligation Intermediates in Human Disease

It is evident from the studies described above that DNA ligase deficiency causes genomic instability, which in turn increases susceptibility to cancer. During DNA replication and DNA excision repair, it is possible that DNA ligases will encounter DNA nicks with mismatches because of misincorporation by a DNA polymerase. DNA joining in vitro by DNA ligases I and III is inhibited by 3′ mismatches at DNA nicks (106), providing an opportunity to remove and then replace the mismatched nucleotide. Thus, DNA ligases can actively contribute to the fidelity of DNA replication and repair. DNA ligases may also encounter damaged DNA termini, such as those generated when oxygen free radicals cause a DNA strand break. Recent work by West and colleagues (9) has implicated an intermediate of the DNA ligation reaction, the DNA adenylate, in the pathology of certain human neurodegenerative diseases. Specifically, they have shown that the DNA adenylate intermediate accumulates when DNA ligase IIIα attempts to join DNA strand breaks with modified termini, such as those generated by oxygen free radicals, and following abasic site cleavage by the major human AP endonuclease (9, 11). Under normal circumstances, the adenylate group is removed by aprataxin (9), which is linked to DNA ligase IIIα by a physical interaction with its partner XRCC1 (107). In addition, aprataxin also interacts with XRCC4, the partner protein of DNA ligase IV (108). The gene encoding aprataxin is mutated in the inherited human neurological disorder ataxia oculomotor apraxia-1, characterized by early-onset cerebellar ataxia, oculomotor apraxia, and late peripheral neuropathy (12, 13). Because aprataxin appears to be the only enzyme that can remove an adenylate group from a 5′ phosphate terminus, the loss of aprataxin is particularly deleterious in postmitotic cells, such as those of the adult nervous system, owing to their more limited repertoire of DNA repair pathways compared with proliferating cells. Attempts to ligate DNA breaks with damaged termini in the absence of aprataxin results in the accumulation of the unrepairable and cytotoxic DNA adenylate in postmitotic neuronal cells (9).

DNA Ligase I

Several DNA ligase I-interacting proteins have been identified by DNA ligase I affinity chromatography (45, 109, 110). Among these interacting proteins are two DNA-sliding clamps, the homotrimeric proliferating cell nuclear antigen (PCNA) and the heterotrimeric hRad9-hRad1-hHus1 (9-1-1) complex, which are involved in DNA replication and cell cycle checkpoints, respectively. The interaction with PCNA but not 9-1-1 is primarily mediated by a conserved PCNA-binding motif, or PIP box, at the N terminus of DNA ligase I (45, 55). This same region is required for the targeting of DNA ligase I to replication factories (Figure 4). The interaction with PCNA is required for localization of DNA ligase I at the sites of DNA replication because inactivation of PCNA binding also abolishes targeting to replication factories (41). As with other PCNA interacting proteins, the PIP box of human DNA ligase I and yeast Cdc9 DNA ligase binds to the interdomain connector loop (IDCL) of PCNA and, at least for Cdc9, involves formation of an additional β−zipper structure between residues adjacent to the PIP box and residues within the C-terminal domain of PCNA (111). This interaction is necessary for complex formation both in the absence of DNA and when PCNA is topologically linked to DNA. The DNA ligase I–PCNA complex formed on DNA contains one molecule of DNA ligase I per PCNA trimer, suggesting that the binding of DNA ligase I to one PCNA monomer occludes the IDCL-binding sites on the other two PCNA subunits of the trimer (109). Inactivation of the DNA ligase I PIP box not only abolishes the targeting of DNA ligase I to replication foci but also the complementation of the Okazaki fragment joining and long-patch base excision repair defects in the DNA ligase I-deficient cell line 46BR.1G1 (41, 55).

The similarity in the size and shape of the rings formed by PCNA and the catalytic fragment of DNA ligase I suggested that the PCNA ring encircling DNA may facilitate the transition of DNA ligase I from an extended conformation into the ring shape via interactions between one of the surfaces of the PCNA ring and regions within the catalytic fragment of DNA ligase I (26). In accord with this idea, a direct interaction between the DBD of DNA ligase I and PCNA was detected (112). Thus, it appears that DNA ligase I initially docks with a PCNA trimer via the interaction between the DNA ligase I PIP box and the IDCL of one of the PCNA monomers (Figure 7a). We envision that the region adjacent to the PIP box is flexible, allowing the catalytic fragment to engage both the PCNA ring and DNA. Despite these protein-protein interaction interfaces, a large molar excess of PCNA is required for significant stimulation of DNA ligase I activity (113). In contrast, the heterotrimeric 9-1-1 clamp is much more effective at stimulating DNA ligase I catalytic activity (114, 115). Because the heterotrimeric PCNA clamp of Sulfolbus solfataricus also stimulates its cognate DNA ligase (116), the ability to stimulate DNA ligation appears to be a common feature of heterotrimeric clamps. Although the mechanism by which the heterotrimeric clamps stimulate DNA joining has not been elucidated, it does not require the topological linkage of the clamp to DNA (115). Interactions between heterotrimeric DNA sliding clamps and the DBD domain of DNA ligases have varying degrees of specificity. The Sulfolobus DNA ligase preferentially interacts with one subunit of the heterotrimeric clamp, whereas human DNA ligase I interacts similarly with each of the 9-1-1 subunits (45, 116). Small-angle X-ray scattering studies with the Sulfolobus ligase and PCNA revealed that the DNA ligase remains in an extended conformation when it is complexed with the heterotrimeric PCNA in the absence of DNA (112). Thus, when the DBD domain of DNA ligase I engages the PCNA ring after the initial PIP box-mediated docking, the catalytic region may still be in an extended conformation (Figure 7b). In this scenario, the conversion of the catalytic fragment presumably involves additional contacts with nicked DNA (Figure 7c).

The PCNA and 9-1-1 clamps are loaded onto DNA in an ATP-dependent reaction catalyzed by replication factorC (RFC) and by an alternative RFC complex containing the human homolog of the Schizosaccharomyces pombe Rad17 protein (hRad17), respectively (117). DNA ligase I interacts with the unique large subunits, RFC p140 and hRad17, of these alternative RFC heteropentamers, which share in common the four small subunits (45, 110). RFC inhibits DNA ligase I, although preincubation of RFC with PCNA alleviates this inhibition, provided that DNA ligase I has a functional PIP box (110). Similar physical and functional interactions occur among the S. cerevisiae homologs, suggesting that this interplay is biologically relevant (111). On the basis of these results, DNA ligase I may be held in an inactive complex with RFC before it is delivered to a PCNA trimer located at a DNA nick that has already been processed by DNA replication and/or excision repair activities. Alternatively, DNA ligase I may interact with a complex of RFC and PCNA left behind at the DNA nick after gap-filling synthesis and end processing. Finally, it is possible that ligation by DNA ligase I is the signal for unloading of the PCNA trimer, thereby recycling PCNA for subsequent Okazaki fragments or repair events. In contrast to the inhibitory effect of RFC, the checkpoint clamp loaders (hRad17-RFC and S. cerevisiae Rad24-RFC) stimulate DNA joining by the replicative DNA ligase of humans and yeast (45). Further work is needed to elucidate the functional and biological relevance of the interactions between DNA ligase I and the clamp loaders.

In addition to interacting with DNA sliding clamps and clamp loaders, DNA ligase I also interacts with DNA polymerase β within a base excision repair complex that was purified from bovine testis (118). This appears to be an alternative version of the major DNA ligase IIIα-dependent short-patch BER pathway that may operate in specific tissues and/or cell types.

DNA Ligase III

As mentioned above, the LIG3 gene encodes four different polypeptides (60, 62). To date, no unique protein partners have been identified for the germ cell-specific isoform, DNA ligase IIIβ, which lacks the C-terminal BRCT motif of polypeptides encoded by the ubiquitously expressed DNA ligase IIIα mRNA (62). The mitochondrial and nuclear versions of DNA ligase IIIα are distinguished by an N-terminal mitochondrial targeting sequence that is proteolytically removed during transport into mitochondria (60). Although XRCC1 forms a complex with nuclear DNA ligase IIIα (64), it is absent from mitochondria (70). Within these organelles, DNA ligase IIIα interacts with DNA polymerase γ, the mitochondrial DNA polymerase (119). This interaction, which presumably integrates DNA ligase IIIα into mitochondrial DNA replication and repair, is not mediated by the BRCT domain of DNA ligase IIIα but instead involves the catalytic region (119).

The discovery of the interaction between XRCC1 and nuclear DNA ligase IIIα (64), which occurs between the C-terminal BRCT domains of these proteins (Figure 8), provided the first clues as to the cellular functions of the polypeptides encoded by the LIG3 gene. Moreover, the structure of the C-terminal BRCT domain of XRCC1 was the first of its type to be determined and has provided the framework for studying the function of this domain in other proteins, including DNA repair and cell cycle regulatory proteins as well as the amino acid changes in the BRCT domain of the BRCA1 protein that are linked to breast cancer (120). In addition to DNA ligase IIIα, XRCC1 interacts with several other proteins involved in base excision and single-strand break repair (Figure 8), suggesting that XRCC1 may act as a scaffold for the assembly of multiprotein repair complexes. By interacting with the hOGG1 DNA glycosylase (121),AP endonuclease (122), and DNA polymerase β (123) in addition to DNA ligase IIIα, XRCC1 potentially contributes at every step of the short-patch base excision repair pathway in which a single nucleotide is inserted in place of the damaged base (123). In vivo XRCC1 is recruited to single-strand breaks by an interaction with automodified PARP1 (75). Polynucleotide kinase (PNK) (124), aprataxin (107), and aprataxin- and PNK-like factor (APLF) (125, 126) share a similar fork head-associated (FHA) domain that interacts with phosphorylated XRCC1. All of these proteins appear to contribute to the processing of DNA termini during single-strand break repair.

Similar to XRCC1, DNA ligase IIIα preferentially interacts with automodified PARP1 and is recruited to DNA single-strand breaks (76). As mentioned above, DNA ligase IIIα is associated with aprataxin, a DNA repair protein that protects against neurodegeneration (9, 107). DNA ligase IIIα also directly interacts with TDP1, a protein involved in the repair of covalent toposiomerase I-DNA adducts and DNA single-strand breaks generated by reactive oxygen species (127). Notably, TDP1 is defective in the neurodegenerative disease spinocerebellar ataxia with axonal neuropathy-1 (127), a second example emphasizing the importance of the DNA ligase IIIα-dependent repair of DNA single-strand breaks generated by reactive oxygen species in neuronal cells. In addition, DNA ligase IIIα interacts directly with the NEIL1 and NEIL2DNA glycosylases that repair oxidized bases in a subpathway of BER, which is dependent upon PNK but not APE1 (128, 129). This complex network of protein-protein interactions involving nuclear DNA ligase IIIα and XRCC1 raises many questions because XRCC1 has repair functions that are independent of DNA ligase IIIα (69). Although XRCC1 and DNA ligase IIIα participate together in different nuclear DNA repair pathways, including single-strand break repair and BER subpathways, it appears likely that different overlapping subsets of the interacting proteins will define the various repair pathways. Indeed, there is evidence that the DNA ligase IIIα-XRCC1 complex is present in different larger multiprotein complexes (107), indicating that XRCC1 and DNA ligase IIIα can simultaneously interact with several repair proteins and suggesting that some of the repair transactions may be carried out by preexisting complexes. Alternatively, repair reactions may be coordinated by sequential handovers of intermediates between repair proteins that were recruited individually to the sites of damage. In particular, this is likely to be the case for PNK, aprataxin, and APLF, which are all involved in cleaning termini at DNA breaks but have overlapping binding sites onXRCC1 (107, 124–126) (Figure 8).

DNA Ligase IV

As mentioned above, DNA ligase IV is unstable in the absence of its partner protein XRCC4 (88). Although the C-terminal extension from the catalytic region of DNA ligase IV contains two adjacent BRCT domains (Figure 6), it is the linker sequence between the BRCT domains that interacts with XRCC4 (130). Interestingly, XRCC4 resembles structural maintenance of chromosome proteins in that it has an N-terminal globular head and a C-terminal coiled-coil domain (131). Although XRCC4 alone forms dimers and tetramers, structural studies indicate that a single DNA ligase IV polypeptide interacts with the dimeric form in the DNA ligase IV-XRCC4 complex (28, 131). Within the XRCC4 dimer, there are interactions between the globular head domain and the regions of the coiled-coil tail adjacent to the head domains (131). The linker region between the BRCT domains of DNA ligase IV binds asymmetrically to the same region of the two coiled coils and induces a large change in the arrangement of the coiled coils (28, 132). To date, there are no available structures of fragments of DNA ligase IV containing the catalytic region, so the position of the catalytic domains relative to the C-terminal BRCT domains and XRCC4 is not known.

Unlike DNA ligase I and nuclear DNA ligase IIIα, which participate in several DNA repair pathways, DNA ligase IV appears to function only in NHEJ, except in specialized cells of the immune system where it also completes V(D)J recombination (91). As shown in Figure 9a, there are physical and functional interactions between DNA ligase IV-XRCC4 and both the Ku proteins and the catalytic subunit of DNA-dependent protein kinase (DNA PK_cs) in NHEJ complexes (133–136). In humans, DNA PK mediates end bridging or synapsis of two DNAs (137), an activity that promotes intermolecular joining by DNA ligase IV-XRCC4 (134). The ends at most DNA double-strand breaks generated in vivo will require processing prior to ligation. In Figure 9b, interactions between DNA ligase IV-XRCC4 and various end-processing factors are shown. Notably, DNA ligase IV-XRCC4 associates with the Pol X family polymerases, Pol μ and Pol λ, as well as with terminal transferase (138, 139). These versatile enzymes fill in the short gaps generated during DNA end alignment and processing. In some circumstances, they add nontemplated nucleotides, suggesting that changes in nucleotide sequence at the break site are tolerated as long as the ends are joined. The joining of incompatible DNA ends has been reconstituted with these purified proteins plus Artemis, a nuclease that interacts with DNA PK_cs (138, 140). DNA ligase IV-XRCC4 contributes to this activity by joining two aligned DNA ends having a short gap in one strand or mismatched termini (84). Interestingly, the recently identified NHEJ factor Cernunnos/XLF, which interacts with XRCC4 (141, 142), stimulates the joining of incompatible DNA ends by DNA ligase IV-XRCC4 (143). The DNA ligase IV-XRCC4 complex also interacts with the end-processing activities, PNK (144), and aprataxin (108).

In S. cerevisiae, Dnl4-Lif1, which is the homolog of human DNA ligase IV-XRCC4, interacts with yeast Ku, and as with the homologous human proteins, yeast Ku recruits Dnl4-Lif1 to DNA ends in vitro (97, 145). The major difference between the primary NHEJ pathways of humans and S. cerevisae is the absence of a yeast homolog of DNA PK_cs. In the yeast pathway, the Rad50/Mre11/Xrs2 complex is the end-bridging factor (146). Rad50/Mre11/Xrs2 interacts with and specifically stimulates intermolecular joining by Dnl4/Lif1 (146). At physiological salt concentrations, the joining of linear DNA molecules with DNA ends by Rad50/Mre11/Xrs2, Dnl4/Lif1, and Ku is dependent upon pair-wise, species-specific protein-protein interactions among these factors (Figure 10a). In addition, as shown in Figure 10b, Dnl4/Lif1 interacts with the processing activities, Pol4, and FEN-1 (Rad27) (147), as well as with Nej1, an ortholog of human Cernunnos/XLF (148–150).

Although DNA ligase IV/XRCC4 catalyzes the last step in the major NHEJ pathway in humans, there is emerging evidence from in vitro and in vivo studies that this complex acts at earlier stages in this repair pathway and plays a key structural role in the assembly of functional NHEJ complexes. Studies by the Chu lab (151) with fractionated cell extracts have shown that DNA ligase IV/XRCC4 is required for processing of DNA molecules with incompatible ends. This implies that DNA ligase IV/XRCC4 acts prior to and is necessary for the subsequent recruitment of end-processing activities, providing an explanation as to why DNA molecules with cohesive ligatable ends are predominantly joined without processing (152). Although Ku is required for the recruitment of DNA ligase IV/XRCC4 to double-strand breaks, there are contradictory reports as to the role DNA PK_cs in this recruitment (153, 154). In yeast, the recruitment of Dnl4/Lif1 is dependent upon Ku but not Rad50/Mre11/Xrs2 (97). Interestingly, the binding of Ku to in vivo double-strand breaks is dynamic (97, 154) and is stabilized, at least in yeast, by Dnl4/Lif1 (97). Regardless of the DNA PK_cs contribution, it is evident that the DNA ligase that completes NHEJ has an earlier role in engaging the ends of the DNAs to be joined. The interaction of either DNA PK_cs or DNA ligase IV/XRCC4 with Ku at a DNA end results in the translocation of Ku along the DNA away from the end (133, 155). This presents a problem because Ku is a ring that slides onto the DNA end (156), so ligation is likely to result in the trapping of Ku molecules on the DNA. Given the intertwining of the subunits of the Ku ring (156), a proteolytic mechanism, perhaps triggered by the ligation step of NHEJ, may be required to remove Ku from the ligated DNA.