A crucial aspect of NHEJ must be the ordered and regulated assembly of the ligation machinery at a break, given that the Ku, MRX, and DNA ligase IV complexes are too large to all be coresident at an extreme DSB terminus and that both MRX and Ku have NHEJ-independent functions at DNA ends. What specific protein-protein interactions are required, and how do they facilitate Dnl4-catalyzed ligation of the DSB? To shed light on these questions, we screened a nearly comprehensive set of potential interactions and identified three that could help coordinate NHEJ (Fig. ). Functional analysis of the Yku80 and Xrs2 domains, predicted to interact with DNA ligase IV, confirmed their importance to NHEJ in vivo and revealed that they are substantially redundant with respect to NHEJ function. Specifically, clear mutant defects demonstrate that the extreme C terminus of Yku80 and the FHA domain of Xrs2 are required for NHEJ, with a strict separation from the telomere, checkpoint, and recombination functions of these proteins. Below we consider how these protein domains might act to facilitate NHEJ.
Our two-hybrid and mutational analyses of the Yku80 C terminus suggest that it has an important role in binding and perhaps positioning or stimulating Dnl4. All experiments, including our inability to demonstrate such binding in vitro to date, indicate that this interaction is weak, however. Thus, we cannot rule out the possibility that the observed NHEJ function of this Yku80 region is something other than direct Dnl4 binding. Alternatively, the weakness of the interaction may simply reflect its greater relevance in the context of DSB-bound proteins. Indeed, Dnl4 binding by the Yku80 C terminus is consistent with the known structure of DSB-bound Ku. Human Ku, and presumably yeast Ku, forms a ring that slides onto DNA in only one orientation with respect to the end (Fig. ) (42
). The crystallized Ku protein lacked the C-terminal tail of Ku80, demonstrating that this region is not required for end binding. However, it is clear that the C terminus of Ku80, and not Ku70, will be oriented to face the DSB, putting it in position to promote ligation.
FIG. 6. The NHEJ critical region of theYku80 C terminus. (A) The crystal structure of the human Ku70/80 heterodimer (light gray) bound to DNA (dark gray) reveals that the C terminus of Ku80 is oriented toward the DSB terminus. The C-terminal residue of the partial (more ...)
To explore this in more detail, we compared multiple Ku80 sequences, available solution structures of the human Ku80 C terminus (18
), and secondary structural predictions using PHD/PROF (32
). This analysis revealed three Ku80 C-terminal types in eukaryotic cells (Fig. ). The first type, exemplified by human Ku80, shows an extended region terminating in an α-helical bundle, as seen by nuclear magnetic resonance (NMR), and a 12-amino-acid C-terminal region that interacts with DNA-PKcs (16
). The second type, exemplified by the fungus Magnaporthe grisea
, retains the same apparent helical-bundle structure but lacks the last DNA-PKcs-interacting region. Indeed, this helical bundle is conserved in most eukaryotes (18
), but only organisms with DNA-PKcs homologues contain the last region (10
). Intriguingly, the third class, including Saccharomyces cerevisiae
and its closest relatives, does not even harbor the helical bundle.
Given this divergence of the Ku80 C terminus, it is not immediately clear if it functions analogously in all organisms. However, the Yku80 C terminus does contain a single putative α-helix that corresponds precisely to the NHEJ critical region (amino acids 608 to 615). Strikingly, this helix may be conserved just N-terminal to the M. grisea
helical bundle (Fig. ). Moreover, the DNA-PKcs-interacting region at the extreme C terminus of human Ku80 is also predicted to fold as a very similar amphipathic α-helix (Fig. ) (18
), although, intriguingly, this was unstructured insolution NMR analyses in the absence of DNA-PKcs. Thus,despite their differences, all three Ku80 configurations present a similar potential protein interaction motif on the DSB side of Ku that could act by partner-induced formation and sequestration of the hydrophobic face of an amphipathic helix. This would explain our finding that both yeast NHEJ and the Yku80-Dnl4 two-hybrid interaction are dependent on three highly conserved leucines, since these would all cluster on one helical face (Fig. ). The especially severe defect of L612P mutation is also consistent with this model (see Table S1 in the supplemental material). It is not clear whether this pattern of putative Ku80 helices represents convergent or divergent evolution, but it is intriguing that a Ku80-DNA-PKcs-type interaction motif is also present in Nbs1 and ATRIP, where it mediates interactions with ATM and ATR, respectively (13
Previous work identified an in vitro interaction between Xrs2 and Lif1 (5
). Our studies indicate that the region encompassing the FHA domain of Xrs2 and the C terminus of Lif1 mediates this interaction. We further show its functional importance for NHEJ in vivo. It is most striking that severe NHEJ inhibition requires disruption of both this interaction between Xrs2 and Lif1 and the proposed interaction between Yku80 and Dnl4. Why might two interactions with DNA ligase IV be required? It is possible that multiple interactions are needed to accommodate the variety of DSB end configurations presented to the NHEJ machinery. Indeed, the fact that imprecise NHEJ is more severely affected by the corresponding Yku80 and Xrs2 mutations than simple religation NHEJ implies that these protein regions are especially necessary for more-complex repair events. Another possibility is that redundant binding of Dnl4 increases the kinetics and efficiency of NHEJ, which would be especially important if DNA ligase IV is the rate-limiting step. This two-contact model is also consistent with published data characterizing an inefficient microhomology-mediated end-joining (MMEJ) mechanism, which involves MRX and DNA ligase IV but not Ku (27
). The Xrs2-Lif1 interaction can account for this Ku-independent role of Dnl4, while the absence of the Ku interaction could explain at least in part the inefficiency of MMEJ.
Several recent studies have addressed the role of the Xrs2/Nbs1 FHA domain in the cell. In mammalian cells, the Nbs1 FHA domain is important for nuclear focus formation and checkpoint signaling, but its role in DNA repair is unclear (38
). In yeast, no definitive role for the Xrs2 FHA domain in DSB repair has been identified, and indeed, several recent studies have revealed no phenotype for xrs2
mutants lacking this domain (35
). Our work identifies an important NHEJ role for the Xrs2 region containing its FHA domain, but this role is robustly apparent only in the Yku80 mutant background. FHA domains bind with high specificity to phosphothreonine residues (12
). We do not yet know if a similar function of the Xrs2 amino terminus mediates its interaction with Lif1, but it should be noted that the FHA domain of polynucleotide kinase 3′ phosphatase interacts with the C terminus of XRCC4 (mammalian homologue of Lif1) in a manner dependent on phosphorylation of XRCC4 by CK2 (23
Importantly, the Yku80 C terminus and Xrs2 FHA domains cannot account for the full role of Ku and MRX in NHEJ, because both the yku80
Δ605-629 and xrs2
ΔFHA mutants retain substantial NHEJ activity, unlike the complete gene deletion mutants. This may indicate that MRX and Ku both serve other essential functions in NHEJ. Specifically, Ku may be required for protecting ends from degradation, and MRX may be required to tether the DNA ends together (11
). A nonexclusive possibility is that the putative Mre11-Yku80 interaction detected in our screen, and perhaps other interactions, may help drive formation of an NHEJ repairosome that contains both MRX and Ku. Indeed, although we have identified many protein-protein interactions involved in NHEJ, there are certainly additional interactions we did not detect. Some may be realized only during NHEJ, either because they are DNA dependent or because they require context-dependent modification of proteins. Moreover, the two-hybrid method has a sensitivity threshold that is biased toward stronger interactions. In this regard, Tseng and Tomkinson described an interaction between Dnl4/Lif1 and Pol4 in vitro (40
) that was not recovered in our screen. Further biochemical work will be required to fully understand the extent of interactions between the various NHEJ proteins.