Perinuclear tethers license telomeric DSBs for a broad kinesin- and NPC-dependent DNA repair process

DNA double-strand breaks (DSBs) are often targeted to nuclear pore complexes (NPCs) for repair. How targeting is achieved and the DNA repair pathways involved in this process remain unclear. Here, we show that the kinesin-14 motor protein complex (Cik1–Kar3) cooperates with chromatin remodellers to mediate interactions between subtelomeric DSBs and the Nup84 nuclear pore complex to ensure cell survival via break-induced replication (BIR), an error-prone DNA repair process. Insertion of a DNA zip code near the subtelomeric DSB site artificially targets it to NPCs hyperactivating this repair mechanism. Kinesin-14 and Nup84 mediate BIR-dependent repair at non-telomeric DSBs whereas perinuclear telomere tethers are only required for telomeric BIR. Furthermore, kinesin-14 plays a critical role in telomerase-independent telomere maintenance. Thus, we uncover roles for kinesin and NPCs in DNA repair by BIR and reveal that perinuclear telomere anchors license subtelomeric DSBs for this error-prone DNA repair mechanism. Damaged DNA is often targeted to nuclear pore complexes for repair. Here, the authors show that kinesin-14 mediates this process ensuring error-prone repair, while perinuclear telomere attachment licenses damaged telomeric loci for this repair and kinesin-14 blocks senescence in the absence of telomerase.

T he DNA replication machinery cannot copy the ends of linear chromosomes. To prevent this limitation from leading to the rapid erosion of chromosomes over cycles of DNA replication and cell division, the ends of linear chromosomes consist of telomeres, or repetitive DNA sequences whose length can be maintained by the enzyme telomerase 1,2 . In addition, telomeres act as recruitment sites for chromosomal complexes that then spread and establish silent chromatin or heterochromatin structures within neighbouring regions known as subtelomeres 3,4 .
DNA double-strand breaks (DSBs) within heterochromatic repetitive DNA constitute a great risk to eukaryotic genome stability. In yeast, DSBs within the highly repetitive ribosomal RNA genes (rDNA) can lead to unequal sister chromatid exchange events, compromise chromosome stability and shorten cellular lifespan [4][5][6][7] . DSBs within rDNA repeats were found to transiently exit the Rad52/recombination-repressive environment of the nucleolus in which they typically reside to undergo repair without compromising genome integrity 8,9 . In flies, heterochromatic DSBs exit their Rad51/recombinationrepressive subnuclear domain to undergo repair 10 . Similarly, heterochromatic DSBs were found to relocate outside their domain for repair in cultured human cells 11 . Therefore, DSBs within heterochromatic repetitive DNAs can exit their nuclear domain to undergo repair.
DSBs escaping their repressive nuclear environment may also be targeted to repair-promoting nuclear regions, such as specific locations along the nuclear envelope. In yeast, these locations are the nuclear pore subcomplex Nup84 (composed of seven entities, including Nup84 and Nup145) as well as the Mps3 inner nuclear membrane protein [12][13][14][15] . In fact, DSBs in cells without clear donor DNA sites for repair relocate at least transiently to Nup84 or Mps3 proteins [12][13][14][15] .
All 32 telomeres of budding yeast are arranged in a handful of clusters at certain points along the nuclear envelope 16 . These telomeric clusters do not co-localize with nuclear pore complexes 17 . Intriguingly, the ability to survive a subtelomeric DSB is dependent on Nup84 as well as a number of factors known to anchor telomere clusters to the nuclear envelope 12,13,[18][19][20][21][22] . However, the relationship between perinuclear telomere tethers and nuclear pore subcomplexes in subtelomeric DNA repair remains unknown 12,23 . In addition, it is unclear whether survival is achieved through bona fide DNA repair processes such as non-homologous end joining (NHEJ) or rather via genome patching mechanisms that may help propagate a compromised genome [24][25][26] . Moreover, how DSB sites move within the nucleus is unclear.
Here, we uncover a role for the kinesin-14 motor protein complex in the transient relocation of telomeric and nontelomeric DSBs to Nup84 and reveal that this function promotes break-induced replication (BIR), a recombination-based errorprone DNA repair mechanism. Perinuclear telomere tethers such as cohibin are required for telomeric DSBs to utilize this broader DNA repair process, which also relies on the DNA damage response, chromatin remodelling, microtubules and Mps3. We also find that kinesin-14 counteracts senescence by ensuring efficient recombination-based telomere maintenance in the absence of telomerase. Overall, our work identifies specific roles for motor and nuclear envelope proteins in the promotion of a broad and error-prone DNA repair mechanism while having potentially critical implications for our understanding of genome instability-linked diseases including cancer.

Results
Genetic analysis of subtelomeric DSB repair. To gain insight into the mechanisms by which DNA breaks in subtelomeric regions are repaired, we used cells in which an inducible DSB can be triggered at either a subtelomeric or a more internal control region on the left arm of chromosome XI (Chr. XI; Fig. 1a) 22,27 . In both of the cases, we introduced galactose-inducible I-SceI endonuclease in cells harbouring a URA3 marker gene flanked by two inverted I-SceI cut sites, which enabled us to confirm successful DSB induction by culturing cells on selective media and monitoring for URA3 reporter loss ( Fig. 1a; Supplementary  Fig. 1). Deletion of the Nup84 subcomplex components Nup84 or Nup145, or of the perinuclear telomere-tethering protein Esc1, significantly decreased survival of the subtelomeric DSB, as expected ( Fig. 1b; Supplementary Table 1) 21,22,28 . Disruption of the telomere-tethering complex cohibin (composed of Lrs4 and Csm1 proteins) alone or in combination with Esc1 deletion, similarly compromised subtelomeric DSB survival (Fig. 1b) 6,8 . Combining the deletion of Nup84 with the disruption of Esc1 or cohibin did not further lower survival rates indicating that all of these factors likely operate within the same pathway (Fig. 1b). Importantly, heh1D or nur1D cells, which retain perinuclear telomere anchoring, did not exhibit significant reductions in DSB survival rates (Fig. 1b) 6 . Since telomere clusters do not generally co-localize with nuclear pore complexes at the nuclear periphery, we wondered whether motor proteins might target subtelomeric DSBs to the nuclear pore complexes 16 . Interestingly, genes encoding for cohibin subunits exhibit negative genetic interactions with Cik1 and Kar3, which localize to the nucleus and together constitute the kinesin-14 motor protein  showing how different PCR products amplified across the DSB site after induction/survival indicate different repair pathways. Survivors were initially analysed and categorized into BIR (no product), NHEJ (B0.7 kb product) or incomplete DSB (B2 kb product; very rare). All BIR survivors were subsequently analysed using a generic telomere AC 1-3 primer with a small fraction of survivors yielding products. All PCR products were amplified using their respective primers as illustrated (P1* used for zip/control sequence strains for initial analysis). (e) EtBr-stained CHEF gel showing changes to the size of Chr. XI in wild-type BIR survivors relative to uninduced and NHEJ survivors. (f-h) Compiled subtelomeric or internal DSB survival assays reveal that survival of a subtelomeric, but not an internal, DSB located on the left arm of Chr. XI is dependent on cohibin/Esc1, Cik1 and Nup84 proteins (mean ± s.d.; N ¼ 5). P values (t-test) comparing mutants to wild-type cells are immediately above bars with additional statistical comparisons as shown. (f) Disruption of telomeric SIR proteins increases Rad52-dependent survival of a subtelomeric DSB. (g) Deletion of the NHEJ factors Ku70/Dnl4 or SIR proteins, but not telomere tethers, motor proteins nor NPC components, decreases survival of an internal DSB. (h) Disruption of the Slx5-Slx8 complex compromises both subtelomeric and internal DSB survival. complex [29][30][31][32][33][34][35] . Indeed, disruption of kinesin-14 alone or in combination with Lrs4 deletion, decreased subtelomeric DSB survival to levels similar to those seen in cells deficient in cohibin or Nup84 (Fig. 1b). Kar3-deficient cells exhibited qualitatively greater drops in survival than cik1D cells (Fig. 1b). This may be related to the Cik1-independent Kar3 function, which is shared with Vik1 and Cin8, at the mitotic spindle poles 36  DSB survival (Fig. 1b). In addition, deletion of the spindle checkpoint protein Bub2 had no significant effect on DSB survival (Fig. 1b). More importantly, deletion of the homologous recombination factor Rad52, but not the NHEJ factors Dnl4 or Ku70, greatly decreased the subtelomeric DSB survival rates of wild-type and lrs4D cells (Fig. 1b). Many proteins including the Ku complex (composed of Ku70 and Ku80) and telomerase contribute to a perinuclear telomere-tethering process that is somewhat distinct from cohibin-dependent tethering 13,20,37,38 .
Since Ku disruption is ineffectual in our subtelomeric DSB survival assay (Fig. 1b), this points to a survival pathway that is more dependent on cohibin and Esc1 telomere tethers, at least within our experimental system. As Rad52 is implicated in different DNA recombination processes, we asked whether a specific type of recombinational repair was implicated at subtelomeric DSBs. Genomic analysis of wild-type survivors by PCR as well as contour-clamped homogeneous electric field electrophoresis (CHEF) revealed that cells relied primarily on BIR, which is typically engaged when DNA sequences only on one side of the DSB have a homologous template (Fig. 1c-e). In contrast, only a small fraction of cells underwent Dnl4/NHEJ-dependent repair (Fig. 1c,d). In addition, deletion of the BIR factor DNA Pola subunit Pol32, but not control Pole subunit Dpb3, significantly decreased subtelomeric DSB survival to levels similar to those seen in Rad52-deficient cells (Fig. 1b) 39 . Moreover, disruption of Rad52, Lrs4, Cik1 or Nup84 specifically decreased BIR-dependent DSB survival (Fig. 1c,d). Furthermore, we found that deletion of the resection factor Sae2 increased overall DSB survival and the percentage of NHEJ-dependent survivors, consistent with the notion that resection suppresses NHEJ (Fig. 1c) [40][41][42][43] . Thus, cohibin/Esc1 perinuclear telomere tethers, kinesin-14, nuclear pore subcomplexes and DNA resection cooperate to ensure BIRdependent repair of a subtelomeric DSB.
Loss of perinuclear telomere tethering compromises SIR (silent information regulator)-dependent silent chromatin assembly but SIR deletions promoted Rad52-dependent repair, pointing to a dominant SIR-independent role for cohibin/Esc1, kinesin-14 and Nup84 in BIR-dependent subtelomeric DSB repair ( Fig. 1f) 4,6 . This is also consistent with the notion that the complete loss of telomeric silencing observed in SIR-deficient cells promotes access to the DSB repair machinery. In contrast, survival of the internal DSB was greatly dependent on SIR proteins as well as the NHEJ factors Dnl4 and Ku70, but not on Rad52, telomere tethers, motor or nuclear pore proteins ( Fig. 1g; Supplementary Table 2). Deletion of Slx5/Slx8 proteins, which are linked to DNA repair both at the nuclear periphery and more internally within the nucleus, decreased survival rates in both subtelomeric and internal DSB repair assays, as expected (Fig. 1h) 12 . Together, our findings so far indicate that the herein studied BIR-dependent repair mechanism is different from silent chromatin assembly and does not act at a more internal DSB site that is on the same chromosome and is more dependent on NHEJ.
Licensing and microtubules at remodelled subtelomeric DSBs. Next, we asked whether cohibin and kinesin-14 proteins promote the physical interaction of the subtelomeric DSB with the Nup84 nuclear pore subcomplex. Indeed, chromatin crosslinking and immunoprecipitation (ChIP) experiments revealed statistically significant increases in the physical interaction between TEL-XI-L and Myc-tagged Nup84, but not Lrs4, upon DSB induction ( Fig. 2a-d). In addition, the increased interaction between TEL-XI-L and Nup84 was dependent on Lrs4 and Cik1 ( Fig. 2e-g). Thus, we wondered whether Lrs4 and Cik1 perform mechanistically similar roles at telomeres. However, deletion of Lrs4, but not Cik1, compromised silent chromatin as assessed by ChIP analysis using an antibody specific to histone H3 acetylated at lysines 9 and 14 (diAcH3K9/14), a robust mark of open chromatin ( Supplementary Fig. 2) 6,8 . Also, unlike Lrs4 deletion, Cik1 loss did not compromise the localization of a green fluorescent protein (GFP)-tagged TEL-XI-L to the nuclear periphery in either the S or G1 phase of the cell cycle ( Fig. 3a-c). In addition, Nup84 localization to the nuclear periphery was unaffected by the deletion of Lrs4 or Cik1 (Fig. 3d). Moreover, cells deficient in the Nup84 component Nup145 exhibited only a very subtle change in perinuclear telomere localization (Fig. 3a-c). Together, these findings suggest that cohibin, kinesin-14 and Nup84 complexes may exert different, yet intersecting, functions to promote telomere-Nup84 interactions upon DSB induction.
We then asked whether DSB induction promotes kinesin-14 localization to the subtelomeric DSB. Indeed, ChIP analysis revealed that upon DSB induction, there is an Lrs4-dependent increased interaction of Cik1 with the DSB site ( Fig. 4a,b). These results suggest that cohibin mediates the recruitment of kinesin-14 to the DSB site. This may be achieved by the ability of cohibin to ensure perinuclear telomere anchoring or by a more direct ability of the complex to physically recruit kinesin-14 proteins to the DSB site. However, co-IP experiments failed to detect physical interactions between TAP-tagged Lrs4 and Myc-tagged Nup84 before and after DSB induction (Fig. 4c). In contrast, in our positive control cells, interactions between TAP-tagged Heh1 and Myc-tagged Lrs4 were observed in the presence or absence of DSB induction (Fig. 4c). Thus, kinesin-14 displays increased localization to the subtelomeric DSB upon its induction in a manner that is likely dependent on cohibin-dependent perinuclear telomere tethering.
Next, we asked whether the DNA repair processes studied here intersect with various factors, some of which were previously linked to DSB targeting to the nuclear envelope. For example, the DNA damage response mediator Rad9 (53BP1 in mammals) and Swr1-dependent Htz1 (also known as H2A.Z) incorporation were previously reported to promote DSB mobility 15 . Indeed, we found that subtelomeric DSB repair was greatly dependent on Rad9 (Fig. 4d). Survival analysis also revealed that swr1D cells had decreased repair efficiency (Fig. 4d). Repair was similarly decreased in cik1D and cik1D swr1D cells (Fig. 4d). In addition, ChIP analysis indicated that subtelomeric DSB-Nup84 interactions as well as Htz1 incorporation were similarly decreased in swr1D and cik1D swr1D cells (Fig. 4e,f). DSBs relocating closer to the nuclear envelope often exhibit increased physical interactions with the Mps3 inner nuclear membrane protein 13,15 . Indeed, anti-Mps3 ChIP analysis showed that subtelomeric DSB-Mps3 interaction levels are increased in cells lacking Lrs4 or Cik1 ( Fig. 4g-i). The similarity between the lrs4D and cik1D cells in these experiments is likely related to the fact that Cik1 displays Lrs4-dependent increased interaction with the induced DSB site (Fig. 4a,g). We also assessed DSB repair in cells lacking the a-tubulin subunit Tub3 as a way to destabilize microtubules without inducing severe cellular defects linked to chemical-based microtubule perturbations 44 . Interestingly, tub3D cells displayed decreased DSB repair rates similar to those seen in cik1D cells (Fig. 4d). In addition, deletion of Cik1 and Tub3 had a partly additive effect (Fig. 4d). These findings suggest that the kinesin-14 complex and microtubules have intersecting yet partly different functions within this DNA repair process. Taken together, these findings provide key mechanistic insights into subtelomeric BIR-dependent DSB repair. Subtelomeric sites experiencing a DSB are subjected to chromatin remodelling and exhibit increased kinesin-14 levels in a manner that is dependent on perinuclear telomere tethering. Such DSBs associate more closely with Mps3 before being ultimately relocated to Nup84 by ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms8742 kinesin-14, which may perform its actions, in part, through cooperation with microtubules.
Artificial hyperactivation of subtelomeric DSB repair. If targeting the subtelomeric DSB to NPCs is a limiting step in BIRdependent repair, we would suspect that artificial targeting of the DSB site to NPCs may hyperactivate this DNA repair process. Thus, we integrated near the DSB site on TEL-XI-L either a control DNA sequence or a so-called DNA 'zip code' sequence that can artificially target various genetic loci to the NPC (Fig. 5a) [45][46][47] . Spotting assays revealed that the integration of such sequences did not impair the efficiency of DSB induction ( Supplementary Fig. 3). ChIP analysis revealed that the zip code sequence, but not the control sequence, led to increased constitutive interaction between the uninduced break site and Nup84 (Fig. 5b,c). This was matched with a drastic increase in subtelomeric DSB repair mainly through BIR in wild-type zip code cells as compared with wild-type control sequence cells (Fig. 5d). In the absence of Lrs4, Cik1, or control Nup84, the zip code sequence failed to elicit significant increases in constitutive telomere-Nup84 interactions or in subtelomeric DSB repair (Fig. 5b-d). The fact that the DNA zip code only increased repair efficiency in wild-type cells but not in cells lacking Lrs4, Cik1 or Nup84, argues against this increased repair being a nonspecific effect related to the zip code disrupting surrounding chromosomal features. In agreement with this rationale, ChIP analysis revealed that the status of diAcH3K9/K14 open chromatin marks near the DSB was similar in cells harbouring zip code and control sequences ( Supplementary Fig. 4). These   Kinesin-14 and BIR-dependent repair at non-telomeric loci. Next, we asked whether cohibin, kinesin-14 and/or Nup84 are part of a broader BIR process that may be acting beyond telomeres. Therefore, we used a non-telomeric BIR assay in which BIR at an HO-induced break recreates a functional CAN1 gene while removing the HPH hygromycin resistance gene (Fig. 6a) 48 .
In this system, over 99% of cells surviving the break utilize BIR, which can also be monitored by PCR 48 . Deletion of the BIR factor Pol32 greatly reduced BIR, as expected (Fig. 6b). Importantly, disruption of Nup84 or kinesin-14 proteins, but not cohibin, decreased BIR (Fig. 6b). BIR kinetics were also impacted in cells lacking Pol32, Cik1 or Kar3 as revealed by PCR analysis (Fig. 6c).
In addition, Kar3 ATPase domain mutants (kar3-1 and kar3-898) failed to complement kar3D cells and restore full BIR-dependent survival ( Fig. 6d) 36,49,50 . kar3-1 and kar3-898 are two Kar3 motor domain mutants that fail to hydrolyse ATP either in general or upon microtubule binding, respectively 49 . This might explain why BIR is slightly higher with kar3-898 as compared with kar3-1 (Fig. 6d). Anti-Nup84 ChIP analysis revealed that kar3D cells and kar3-1 cells, respectively, display decreased and increased interaction between the induced DSB site and Nup84 (Fig. 6e). We observed this in spite of the fact that both kar3D cells and kar3-1 cells have similarly decreased repair efficiency (Fig. 6d). These results suggest that kinesin-14 targets DSBs to and away from Nup84 via mechanistically different processes. This also indicates that both the relocation to and release from Nup84 are important for repair. We also assessed the role of kinesin-14 and Nup84 in the repair of a DSB known to relocate to the nuclear periphery 12,13 . We used a system in which Chr. III is modified to harbour an HO-inducible DSB site within a leu2 locus (Fig. 6f). Either 5 or 30 kb away from the DSB site, a leu2 segment is also inserted. The 5 kb strain is preferentially repaired by single-strand annealing (SSA) whereas the 30 kb strain undergoes SSA or BIR and is known to relocate to the nuclear periphery for repair (Fig. 6f) 13,50 . Although the repair of both DSBs were dependent on Rad52, only repair of the 30 kb substrate showed a partial dependency on Cik1-Kar3, Nup84 and Htz1 (Fig. 6g). Of note, kinesin-14 was only minimally involved in gene conversion indicating that the observed major impact on BIR systems is not related to unforeseen changes in general DNA donor-based repair processes (Fig. 6h,i) 51 . Together, our findings indicate that kinesin-14 and Nup84 are part of a BIR pathway that can act beyond subtelomeric DSBs. We note that although we found kinesin-14 to mediate repair at three of three different BIR-dependent reporter DSB sites tested, it remains unclear if all BIR-dependent repair, or even if BIR-independent repair, implicates kinesin-14 and NPCs.
Kinesin-14 maintains telomeres and DSB dynamics. BIR is physiologically important for the maintenance of telomeres in the absence of telomerase via a recombination-based mechanism that is akin to the 'alternative lengthening of telomeres' seen in telomerase-negative human cancers 39,[52][53][54] . Furthermore, eroded telomeres in telomerase-negative yeast cells relocate to nuclear pore complexes, suggesting that the dynamics of telomere localization at the nuclear periphery may be critical for the telomerase-independent maintenance of chromosome ends 23 . To test whether kinesin-14 participates in telomere maintenance in cells lacking the telomerase catalytic subunit Est2, meiotic segregants originating from a common diploid strain were propagated by serial dilution over 10 days and cell densities were recorded daily to estimate cell viability (Fig. 6j). Wild-type, cik1D or kar3D cells maintained constant levels of viability over the course of the experiment, whereas est2D cells showed a modest senescence profile manifested by a slight dip in cell viability followed by recovery (Fig. 6j). In contrast, the est2D cik1D and est2D kar3D cultures displayed accelerated senescence for 4 days before slowly recovering cell viability by the end of the experiment (Fig. 6j). This initial accelerated senescence is reminiscent of telomerase-negative rad52D or pol32D cells, although these latter cells typically fail to exhibit recovery 39,52,53 . This is consistent with the fact that the loss of Rad52 or Pol32 decreases BIR-dependent repair more drastically than the loss of kinesin-14. More importantly, early accelerated senescence in kinesin-14-deficient est2D cells correlated with a decreased telomeric repeat signal (Fig. 6k). Taken together, our findings indicate that kinesin-14 plays a critical role in early telomeraseindependent telomere maintenance.  We finally tested the possibility that kinesin-14 ensures that damaged DNA maintains a dynamic character, which is needed to move within the nucleus 55,56 . Repair of both of our internal and subtelomeric BIR DSB sites is dependent on Rad52 and previous reports using mean square displacement (MSD) analysis revealed that YFP-tagged Rad52 closely follows the dynamics of its DSB sites 55,56 . Thus, we examined the dynamics of the induced internal BIR DSB, as well as the subtelomeric DSB sites,  in wild-type and kinesin-14-deficient cells. Deletion of Cik1 or Kar3, or expression of the kar3-1 point mutant, in cells harbouring internal BIR DSBs resulted in greatly decreased Rad52-YFP dynamics (Fig. 7a,b). Non-damaged DNA moves within the nucleus with a diffusion coefficient D of up to B10 À 3 mm 2 s À 1 in yeast, bacteria, flies and mammalian cells 57 . DNA damage would be predicted to increase D beyond this limit. Thus we compared the percentage of cells with D410 À 3 mm 2 s À 1 in wild-type and kinesin-14-deficient cells. Deletion of Kar3 or expression of kar3-1, respectively, decreased this percentage from 31.3% in wild-type cells to 4.4 and 11.4% (Fig. 7b). In our analysis of Rad52-YFP in the subtelomeric DSB system, Cik1 deletion mildly impacted the MSD curve although a decrease from 21.4 to 14.6% of cells with D410 À 3 mm 2 s À 1 was still observed (Fig. 7c). This may reflect the fact that, unlike the internal BIR DSB, the subtelomeric DSB is initially closer to the nuclear envelope and so the impact of kinesin-14 on the dynamics of the latter site may be harder to detect. Thus, kinesin-14 allows BIR DSBs to adopt a dynamic character in the nucleus further highlighting the role of DSB mobility in the BIR-mediated repair processes uncovered here.

Discussion
Our results support a model in which DSBs depend on Swr1/ Htz1-dependent chromatin remodelling and kinesin-14 to transiently interact with the nuclear pore subcomplex Nup84 in a BIR-dependent DNA repair pathway (Fig. 7d). Perinuclear telomere tethering ensures increased kinesin-14 localization to DSBs thereby licensing them for this broader DNA repair process. Our findings also suggest that DSBs may be targeted to Mps3 before being ultimately relocated in a kinesin-14-dependent manner to Nup84. Kinesin-14 mobilizes DSBs to and away from Nup84 in a manner that is independent and dependent of the catalytic activity of kinesin-14, respectively. These functions of kinesin-14 implicate microtubules. Future work should help clarify how kinesin-14 performs mechanistically different roles to cooperate with microtubules and transiently relocate DSBs to Nup84. It is possible that nuclear microtubule tracks are laid on demand as such structures were recently observed in a different context, quiescence 35 . Cik1-Kar3 is a microtubule minus end-directed motor complex that can only walk on microtubules with a combination of an active (Kar3) and inactive (Cik1) leg 58 . We found that the catalytic activity of Kar3 is required only to release DSBs from Nup84. Thus, it is possible that microtubules are laid with the plus end at NPCs in which case kinesin-14 would have to enlist the catalytic activity of another plus-end-directed motor protein complex to mediate DSB targeting to NPCs. Consistent with this rationale, kinesin-14 type complexes were previously proposed to operate within multi-motor ensembles to collectively organize microtubule networks in vivo 59,60 . This arrangement would also be consistent with the fact that microtubule minus ends are relatively static, whereas plus ends are mobile and often engage in so-called dynamic instability, which is defined by cycles of microtubule growth and shortening 61 .
DSBs can exhibit greatly increased mobility, which is thought to promote the homology search during recombinational repair 55,56,62 . These processes may be conserved, as broken DSB ends can exhibit limited or pronounced mobility in mammals [63][64][65][66] . Thus, it will be important to determine whether motor complexes contribute to DSB repair in different organisms including human.
We also demonstrate that recently discovered DNA zip codes can be used to target chromosome ends to nuclear pore complexes in a constitutive manner. Interestingly, this action of  . P values (t-test) comparing control mutants with control wild-type cells are shown immediately above bars. Statistical comparison of zip and control pair are also presented for each of the mutants. Survival analysis was conducted as in Fig. 1c,d. the zip code is dependent on the telomere-tethering complex cohibin. This may indicate that perinuclear telomere tethering facilitates zip code-dependent targeting of the telomere to NPCs. In contrast, as kinesin-14, unlike cohibin, is not required for perinuclear chromosome tethering, zip code function may directly require the kinesin-14 complex. Taken together, our study supports the notion that DNA zip codes can be used to artificially increase interactions between NPC components and silent chromatin domains such as telomeres.
In conclusion, we have uncovered that conserved kinesin-14 and Nup84 proteins cooperate within a BIR pathway that can act on DSBs at different genetic loci. Importantly, BIR-dependent DSB repair is inherently an error-prone process that can compromise overall chromosome integrity. Thus, the herein identified mechanisms may provide DSB survivors with an advantage or disadvantage within a cell population. This may, in turn, be beneficial or deleterious to an organism depending on specific cellular or environmental conditions. Many secondary ARTICLE tumours or resistant cancers emerge from genomic rearrangement in cells exposed to genotoxic therapeutics 67 . It will be important to determine whether error-prone DNA repair processes similar to the one uncovered here play a role in secondary or resistant cancers. In addition, perinuclear telomere tethers have been viewed as guardians of chromosome ends 4,21 . However, our findings suggest that in the presence of a DSB, perinuclear tethers can help propagate compromised genomes through engagement of the BIR pathway.

Methods
Strains and materials. Endogenous genes were deleted or modified with C-terminal epitope tags 6,8 . Cells were grown to log phase, 5 ml of cell culture was centrifuged, washed with sterile water and resuspended in 100 ml LiOAc mix (100 mM LiOAc pH 7.3, 10 mM Tris-Cl pH 8.0, 1 mM EDTA). Fourteen microlitres of precipitated transformation DNA generated via PCR was added to each sample, followed by the addition of 30 ml salmon sperm single-stranded DNA (10 mg ml À 1 ; Invitrogen) and 700 ml PEG (21 g PEG3350, 100 mM LiOAc pH 7.3, 10 mM Tris-Cl pH 8.0, 1 mM EDTA). The transformation mixture was incubated at 30°C in a water bath for 45 min followed by the addition of 100 ml of DMSO to the sample and then cells were heat shocked in a 42°C water bath for 15 min. Cells were pelleted via centrifugation and resuspended in 200 ml SOS mix (1 M sorbitol, 1/3v/v YEP media, 6.5 mM CaCl 2 ). Cells transformed with DNA using a drug selection marker was plated onto YEPD media and replica plated the next day onto their respective drug selection plates. Yeast strains are listed in Supplementary  Table 3. Primers for PCR survivor analysis, ChIP and zip code integration are in Supplementary Table 4. For zip/control sequence integration, primers listed in Supplementary Table 4 and containing either the zip or control sequence were designed and used to amplify a PCR product for transformation following the same protocol used to generate endogenous gene deletions. The NatR cassette gene was used for the selection of integrated strains and the zip/control sequences were B1.4 kb left of the DSB (Fig. 5a). Integration of the zip on the right of the DSB site disrupted DSB induction capacity and so was not used.
Chr. XI DSB assays and survivor analysis. Strains containing the URA3 cassette flanked by two inverted I-SceI cutting sites at different positions along Chr. XI were transformed by the I-SceI expression plasmid pKM97 (refs 22,27). This plasmid harbours the I-SceI ORF under control of a galactose-inducible promoter and induction of DSB was performed by plating on galactose media 22 . All subtelomeric (Figs 1b,c,f,h and 4d) or internal (Fig. 1g,h) DSB survival rates were compiled from at least five platings of two independent pKM97 transformants and compiled wildtype rates are included on the different graphs with t-test statistical analyses comparing either mutant and wild-type strains to each other or indicated mutant strains to each other. A wild-type set was included as an internal reference for every round of mutants plated. For survivor analysis (Figs 1c and 5d), at least 30 survivors were analysed by PCR using primer pairs positioned across the DSB site as in Supplementary Table 4. Survivors identified as BIR were further analysed by PCR using another set of primers (telomeric and more internal on Chr. XI; Fig. 1d) as presented in Supplementary Table 4.
Non-telomeric BIR assay. JRL346-derived strains were grown in XY þ 2% raffinose before plating onto XY þ 2% glucose or XY þ 3% galactose to induce the single DSB 48 . Cells surviving the DSB were replica plated onto Hygromycin B plate to confirm the outcome of repair. The fraction of DSB repair by BIR was calculated by dividing the Hygromycin B sensitive colony-forming units by the initial colonyforming units (cell survival on glucose media). To study the kinetics of BIR, JRL346-derived strains as indicated were grown in XY þ 2% raffinose and 3% galactose was added in log phase cultures to induce the DSB. Time points were taken at 0, 3, 6, 9, 12, 19 and 24 h post induction and genomic DNA was extracted. Semi-quantitative PCR was used to measure the unique BIR repair product using primers P1 and P2 (ref. 48). Gels were imaged using a Bio-Rad Gel Doc XR þ with constant parameters and the resulting images analysed in ImageQuant using volume quantification. Percent repair by BIR was calculated by normalizing the BIR product amplicon signal to the loading control HTA2 amplicon, subtracting the t ¼ 0 background signal, and normalizing to genomic DNA from a clone that had undergone BIR.
Gene conversion assay. Using the TGI354 assay to study ectopic gene conversion, the DSB was induced at the mating type locus (MAT) by galactose-controlled (c) Mobility of subtelomeric (SubTEL) DSBs is shown (N ¼ 84). (d) Model depicting BIR-dependent repair processes at subtelomeric and internal DSBs tested. Swr1-dependent Htz1 incorporation allows kinesin-14 to target BIR-repairable DSBs to the nuclear pore subcomplex Nup84 for repair. Cohibin-dependent perinuclear telomere tethering licenses subtelomeric DSBs for this repair process by promoting kinesin-14 recruitment to damaged DNA. DSBs may increase their association with the inner nuclear membrane protein Mps3 along the nuclear envelope before being ultimately targeted to Nup84. DSB targeting to and release from Nup84 via kinesin-14 is performed via non-catalytic and catalytic functions of the complex, respectively. Kinesin-14 performs its role in part via cooperation with microtubules. Dashed arrows symbolize events that promote DSB targeting to the NPC. Green kinesin-14 and arrows highlight DSB targeting to Nup84. Red kinesin-14 and arrows highlight DSB release from Nup84. expression of the HO endonuclease 51 . This strain lacks donor sequences (HML and HMR) for homologous recombination. The homologous MAT-inc region on Chr. III is used as a donor for gene conversion. Quantification of cell survival after DSB induction by galactose was then conducted.
SSA and BIR assays. We used the YMV45 and YMV2 assays to study SSA and BIR in response to DSBs 68,69 . The DSB was induced by galactose-controlled expression of the HO endonuclease acting at the cleavage site inside the LEU2 open reading frame (leu2::cs) on Chr. III. These cells use a complementary his4::leu2 segment inserted B5 kb (YMV45) or 30 kb (YMV2) away from the HO cleavage site for DSB repair by SSA (YMV45) or a combination of both SSA and BIR (YMV2). Quantification of cell survival after DSB induction by galactose was then conducted.
Spotting assay. Cells were spotted starting with 1.25 Â 10 5 cells per spot and then in 10-fold serial dilutions onto SC, SC þ 5FOA or SC-URA media and incubated at 30°C for 3 days. Cells in which the URA3 reporter is excised by successful galactose-inducible DSB grow on SC þ 5FOA but not SC-URA media.
ChIP. For subtelomeric DSB ChIPs, yeast cultures (50 ml) were first grown in SC-leu with 2% raffinose media and then transferred to either SC-leu with 2% glucose (control) or 2% galactose (DSB induction) for 3 h. Cells were similarly treated for non-telomeric DSB ChIPs (Fig. 6e) except that SC-trp instead of SC-leu media was used. For uninduced/silencing ChIPs, standard YEPD media was used. In both the cases, cells were grown to a D 600 of 1.8 and cross-linked with 1% formaldehyde at room temperature (RT) for 30 min. The reaction was quenched with glycine at a final concentration of 125 mM for 5 min at RT. Cells were washed with cold TBS (20 mM Tris-HCl pH 7.6 and 150 mM NaCl). Cell pellets were resuspended in 400 ml of lysis buffer (50 mM HEPES-KOH pH 7.5, 500 mM NaCl, 1 mM EDTA, 1% Triton X-100, 0.1% sodium deoxycholate, 0.1% SDS, 1 mM PMSF, 1 tablet of complete protease inhibitor (Roche)). Samples were bead-beated with an equal volume of glass beads 2 Â 30 s with intermittent 2 min incubation on ice. Lysates were sonicated 3 Â 20 s at 40% amplitude with intermittent incubations on ice. Two consecutive rounds of centrifugation at 16,000g for 5 and 15 min clarified lysates. One hundred and fifty microlitres of lysate was incubated at 4°C for 2 h with 2 mg of antibody/ sample and further incubated with 15 ml of 50% slurry of pre-washed Protein A-Sepharose beads (GE 17-0870-01) at 4°C for 1 h. Antibodies used were anti-c-Myc 9E10 (Covance MMS-150R), anti-AcK9/AcK14 H3 (Millipore 06-599), anti-Mps3/Nep98 (gift from Toshiya Endo lab, Nagoya University), anti-Nup84 (Abcam ab122883) and anti-Htz1/H2A.Z (Abcam ab4626). Beads were washed 3 Â with lysis buffer, 1 Â with LiCl buffer (10 mM Tris-HCl pH 8.0, 250 mM LiCl, 0.5% NP-40, 0.5% sodium deoxycholate, 1 mM EDTA) and 1 Â with TE (10 mM Tris-HCl pH 8.0, 1 mM EDTA), and eluted with 50 mM Tris-HCl pH 8.0, 10 mM EDTA and 1% SDS. The eluant was incubated at 65°C overnight. For Fig. 4a, extra precautionary measures were taken in keeping samples cold because the fusion protein Cik1-Myc13 was particularly sensitive to post-lysis heat-induced degradation. In this case, samples were bead-beated 4 Â 15 s and lysates sonicated 6 Â 10 s at 40% amplitude with intermittent incubations on ice instead. RNaseA was added at a final concentration of 0.2 mg ml À 1 and incubated at 37°C for 30 min, followed by proteinase K (to 0.2 mg ml À 1 ) and glycogen (to 0.03 mg ml À 1 ) treatment for 2 h at 37°C. Dilutions for IP and input DNA were 1:2 and 1:1,000, respectively. PCR parameters were one cycle of 95°C for 2 min, 55°C for 30 s and 72°C for 1 min, followed by 24 to 31 cycles (same within primer sets, optimized for gel analysis) of 95°C for 30 s, 55°C for 30 s and 72°C for 1 min, and a final step of 72°C for 4 min. Primers and PCR product sizes are listed in Supplementary Table 4. The PCR products were ethidium bromide (EtBr)-stained. For acrylamide gel-based analyses, relative fold enrichment for the Myc IP reactions is defined as the ratio of the Myc-tagged strains values relative to untagged value according to the following calculation: [DSB IP /ACT1 IP ]/[DSB input /ACT1 input ]. The relative fold enrichments for the anti-diAcH3 reactions were similarly generated. For quantitative PCR analyses, equations taking into account dilution factors and ACT1 control signals were used to calculate enrichments at the given genomic sites. For example, [100 Â 2 (adjusted inputsite-Ct (IPsite)) ]/[100 Â 2 (adjusted inputACT1-Ct (IPACT1)) ], where adjusted input ¼ CT input À log 2 1,000.
Live cell imaging. Cells were grown into log phase, pelleted via centrifugation, washed and resuspended in PBS before being mounted onto slides for microscopy. Two-dimensional (2D) images were collected with a Nikon C2 þ Confocal Microscope and processed with NIS-Elements AR (Nikon). Using the LacI-GFP fused tetO system, positions of GFP-marked Chr. XI subtelomeres were determined during the S and G 1 phase of the cell cycle with NIS-Elements AR (Nikon) and representative images were adjusted for background using levels and contrast in Photoshop (Adobe). G 1 phase cells were imaged after incubation with 10 mg ml À 1 alpha-factor (Sigma) at RT for 3 h. Distance of subtelomeric signal to the nuclear envelope was used to assign localization to one of three concentric nuclear zones of equal area.
Immunofluorescence imaging. Cells were fixed with 1% formaldehyde, washed and resuspended in SP buffer (1.2 M sorbitol and 0.1 M KH 2 PO 4 pH 7.0). Cells were spheroplasted for 30 min in a digestion mix (40 mg ml À 1 lyticase (Sigma) and 0.2% 2-mercaptoethanol in SP buffer) and adhered to diagnostic slides pre-coated with 1 mg ml À 1 poly-L-lysine. After blocking in a solution of 1% BSA and 0.1% Triton X-100 in phosphate-buffered saline (PBS) for 1 h, cells were washed with PBS and incubated with goat anti-Nup84 (1:500 in 1% BSA in PBS, Abcam 122883; Fig. 3d) or rabbit anti-Mps3/Nep98 (1:100 in 1% BSA in PBS, gift from Toshiya Endo lab, Nagoya University; Fig. 4i) for 1 h at RT. Cells were washed again and incubated with Alexa Fluor 568 rabbit anti-goat (1:750 in 1% BSA in PBS, Life Technologies A11079; Fig. 3d) or Alexa Fluor 568 goat anti-rabbit (1:500 in 1% BSA in PBS, Life Technologies A11011; Fig. 4i) for 1 h at RT. Cells were washed and counterstained with 1 mg ml À 1 DAPI for 5 min at RT. Samples were washed, Permafluor (Thermo Scientific) was applied as mounting media, and the slides were sealed. The 2D images were collected on a Nikon C2 þ Confocal Microscope and processed with NIS-Elements AR (Nikon). The intensity of Nup84 fluorescence at the nuclear periphery and interior was measured with ImageJ (NIH). Scientific). Single plane images were acquired in the YFP channel every 1.5 s for 180 s (490 nm excitation, 200 ms exposure) while mCherry images were captured every fifth frame (575 nm excitation, 200 ms exposure). Nuclear alignment was performed using Nup49-mCherry frames with a custom MATLAB script while Rad52-YFP was tracked using the SpotTracker plugin in ImageJ to yield x, y coordinates 70 . Final MSD analysis was calculated using a custom MATLAB script. We used the slope (m) of the first five time intervals (1.5-7.5 s) to determine the 2D diffusion coefficient (D) given that D ¼ m/2 Â d, where d is the number of dimensions. The radius of confinement was calculated as Rc ¼ 5/4 Â OMSD calculated from the average MSD value of the last 20 time intervals (121.1-150 s).
Telomerase-independent senescence. Cells were grown to saturation in 10 ml cultures of XY with 2% glucose and diluted to 5 Â 10 5 cells ml À 1 every 24 h. Cell density was measured using a hemocytometer. Samples were taken at days 0, 4 and 8 for telomere length analysis by Southern blotting. Genomic DNA was digested with XhoI and probed with a Y'-TG probe generated from the pYT14 plasmid 1 . The approximate number of generations was calculated on the basis of the initial growth on plates after sporulation (3 days) and the cell density during liquid culturing.