The lack of Lig3 in model organisms like yeast and the lack of Lig3 mutant mammalian cells had limited functional studies of Lig3 in vivo
. The recent discovery that the cellular viability requirement for Lig3 depends on its role in mitochondria 
led to the development of cell lines that are deficient for Lig3 in the nucleus 
, allowing us to address the role of Lig3 in chromosomal translocation formation and alt-NHEJ. Using ZFNs, DSBs were introduced into endogenous loci in these cells without prior integration of reporter substrates; nested PCR allowed the recovery of chromosomal translocation junctions within 60 hours. This system works as efficiently in mouse ES cells as in human ES cells 
. With this approach, we were able to systematically induce and analyze translocations in a variety of ligase deficient backgrounds in mouse cells. Given that they arise by alt-NHEJ even in the presence of the canonical NHEJ 
, translocations provide a good model with which to characterize components of the alt-NHEJ pathway.
Here, we establish that Lig3 is a component of the alt-NHEJ pathway leading to translocations in mouse cells, as Lig3 deficiency leads to a >2-fold decrease in translocation frequency. By examining an extensive number of translocation breakpoint junctions, we observed a redistribution of microhomology with Lig3 loss to that expected by chance, implying that Lig3 favors the use of microhomology during joining. Microhomologies are short, as would be expected from limited end resection which exposes single-strands for annealing 
and from analysis of translocation junctions found in patients 
. The use of short microhomologies in chromosomal rearrangements is further underscored by recent genome-wide analysis of breast cancers 
We also demonstrate that the role of Lig3 in translocation formation is independent of XRCC1, as deletion of the XRCC1-interacting BRCT domain does not affect either translocation frequency or breakpoint junction characteristics. Although Lig3 and XRCC1 have been suggested to work in a complex 
, our results are consistent with recent studies that have differentiated the roles of Lig3 and XRCC1 in DNA damage repair 
. In contrast, deletion of the ZnF domain results in a decrease in translocation frequency. The ZnF domain may promote intermolecular ligation in this context, joining two translocation partners, in agreement with a role for this domain in the ligation of oligomers in vitro 
. The lack of a shift in microhomology use as seen in the absence of Lig3 argues that the ZnF domain is not required for microhomology use in translocations. However, an unusual class of junctions – those with long insertions (>50 bp) – was more prominent in cells expressing this protein. It is conceivable that the deletion of the ZnF domain results in slower kinetics of joining in vivo
, which could allow for longer polymerization giving rise to insertions in a subset of junctions.
Consistent with previous results obtained in mouse cells 
, the loss of XRCC4 or Lig4 increases translocation formation, highlighting once again that the canonical NHEJ ligase suppresses translocations. Lig4–XRCC4 could act as a physical barrier together with Ku to prevent access of Lig3 to DNA ends for translocation formation; alternatively, this complex, together with other canonical NHEJ components, could promote the efficient joining of ends to narrow the kinetic window for translocation formation. Importantly, depletion of Lig4 in nuclear Lig3-deficient cells did not increase the translocation frequency as it did in wild-type cells, such that the relief of the translocation suppression by Lig4–XRCC4 is specifically related to Lig3 access to ends. Thus, Lig3 loss leads to an even greater fold decrease in translocation frequency in Lig4-depleted cells (3.7-fold) than it does in otherwise wild-type cells (2.3-fold). This suggests a more dominant role for Lig3 for translocation formation in the absence of the canonical NHEJ ligase.
We further found that Lig1 depletion in wild-type mouse cells does not have any effect on translocation frequency, whereas Lig1 depletion in nuclear Lig3-deficient cells nearly abolishes translocations. This implies that Lig3 has a primary role in alt-NHEJ resulting in translocations, but that Lig1 can function in the absence of Lig3 as a back-up ligase to provide limited activity, suggesting the existence of at least two alt-NHEJ pathways with these ligases operating in a hierarchy. Recent results examining base excision repair have also indicated that Lig1 and Lig3 can act in a hierarchy, although in this case Lig1 appears to be the primary ligase whereas Lig3 acts as the back-up ligase 
. That these two ligases function in distinct pathways in translocations, as opposed to substituting for each other within one pathway, is supported by the different microhomology distributions: breakpoint junctions formed by Lig3 show a preference for pre-existing microhomology, whereas those formed by Lig1 do not (). We cannot exclude that microhomology is generated by short polymerization (polymerase-generated microhomology) 
that would promote Lig1-dependent joining. Polymerase-generated microhomology could also account for the 0 bp microhomology class of joining events that occur in the presence on Lig3; alternatively, there may not be a strict dependence on microhomology for joining by Lig3. The short polymerization may be template-dependent () or arise by chance in a template-independent manner (not shown). Although polymerase-generated microhomology cannot be scored, the existence of short insertions at translocation breakpoint junctions (either template dependent or independent) provides evidence for polymerization at DNA ends 
Model for the role of the three mammalian DNA ligases in chromosomal translocation formation in mouse cells.
In the end, NHEJ is a DNA ligation process 
, and here we establish the intricate interplay of the three DNA ligases in alt-NHEJ leading to translocations in mouse cells. Lig3 promotes alt-NHEJ, but in its absence Lig1 can also function in this process. The roles for Lig3 and Lig1 contrast with that of Lig4, which suppresses chromosomal translocations.