LEM-3 is a midbody-tethered DNA nuclease that resolves chromatin bridges during cytokinesis

Faithful chromosome segregation and genome maintenance requires the removal of all DNA bridges that physically link chromosomes before cells divide. Using C. elegans embryos we show that the LEM-3/Ankle1 nuclease defines a new genome integrity mechanism by processing DNA bridges right before cells divide. LEM-3 acts at the midbody, the structure where abscission occurs at the end of cytokinesis. LEM-3 localization depends on factors needed for midbody assembly, and LEM-3 accumulation is increased and prolonged when chromatin bridges are trapped at the cleavage plane. LEM-3 locally processes chromatin bridges that arise from incomplete DNA replication, unresolved recombination intermediates or the perturbance of chromosome structure. Proper LEM-3 midbody localization and function is regulated by AIR-2/Aurora B kinase. Strikingly, LEM-3 act cooperatively with the BRC-1/BRCA1 homologous recombination factor to promote genome integrity. These findings provide a molecular basis for the suspected role of the LEM-3 orthologue Ankle1 in human breast cancer.


Introduction
Faithful chromosome segregation requires all connections that physically link chromosomes to be removed before cells divide; failure to do so may lead to the severing of chromosomes, prevent cell division or cause aneuploidy and polyploidization [1][2][3][4] . It is widely known that proteinaceous connections between chromosomes mediated by cohesins must be removed prior to segregation, but so must DNA structures that link chromatids. These DNA connections include DNA repair intermediates, points at which chromatids have become intertwined, and loci that have not fully replicated. Cytologically, connections between separating chromatids can take the form of chromatin bridges or ultrafine DNA bridges 5,6 . The majority of DNA linkages, such as branched recombination intermediates, are 'dissolved' by the combined action of the BLM helicase and Topoisomerase III concomitant with DNA replication, or by the MUS81-EME1 nuclease in G2/M and the GEN1 Holliday junction resolvase during anaphase (for review 7 ). However, remaining chromatin bridges can persist into telophase. During cytokinesis the Aurora B kinase dependent NoCut checkpoint detects chromatin bridges and delays abscission of daughter cells in order to allow for chromatin bridge resolution 4,8 . This delay is achieved through continuous Aurora B activation, while cytokinesis is completed upon kinase inactivation. However, little is known about how chromatin bridges are resolved when the Aurora B kinase delays cytokinesis.
The LEM-3 nuclease was discovered in a genetic screen for DNA repair genes in C. elegans 9 . lem-3 mutants are hypersensitive to ionizing radiation (IR), UV light and DNA cross-linking agents 9 . However, previous studies have not shown how LEM-3 mediates genome stability.
The mammalian orthologue of LEM-3 is known as Ankle1 and is predominantly expressed in hematopoietic tissues. Ankle1-deficient mice are viable and show no detectable defects [10][11][12] Interestingly, non-coding polymorphisms in either the ANKLE1 locus or the neighbouring ABHD8 locus have been associated with increased risk of breast and ovarian cancer in the general population and in carriers of BRCA1 mutations, suggesting that altered expression of ANKLE1 or ABHD8 might contribute to tumourigenesis [13][14][15] .
Here we identify a novel mechanism conferred by LEM-3 nuclease to preserve genome integrity during the final stages of cell division in C. elegans. Our data indicates that LEM-3 can be regulated by the Aurora B kinase AIR-2 and provides a final safeguard to resolve the remaining chromatin bridges at the very end of cytokinesis, by virtue of the recruitment of LEM-3 to the midbody.

LEM-3 acts outside of known DNA repair pathways
C. elegans lem-3 null mutants were reported to proliferate normally but are hypersensitive to a range of DNA damaging agents including ionizing irradiation (IR) which predominantly acts by inducing DNA double-strand breaks 9 . The mammalian orthologue of LEM-3, known as Ankle1, is poorly characterised [10][11][12] . We wished to determine whether LEM-3 acts in previously characterised DNA damage repair pathways, or else defines a new mechanism.
We found that after IR treatment, mutation of lem-3 showed more severe phenotypes when combined with null alleles defective in the three major DNA double-strand break repair modalities: BRC-1 dependent homologous recombination (HR), polymerase Theta (POL-Q) mediated end-joining 16 and LIG-4 dependent non-homologous end-joining 17 (Fig. S1A-S1C).
Taken together, our genetic data indicate that LEM-3 might act in a previously unknown response mechanism.

LEM-3 accumulates at the midpoint of chromatin bridges during cell division
To better understand how LEM-3 helps to maintain genome stability, we assessed LEM-3 localization using an existing transgene carrying a lem-3::YFP translational fusion, as well as a GFP tagged lem-3 derivate, generated by CRISPR-Cas9 mediated genome editing. Both constructs complement the DNA damage hypersensitivity phenotypes conferred by lem-3 null alleles (Fig. 5C) 9 . Consistent with previous findings 9 , LEM-3 appeared excluded from the nucleus, and formed prominent foci of uncertain nature outside the nuclei in the absence of DNA damage (Fig. 1A). To explore potential links between these LEM-3 foci and chromosome segregation, we exposed worms to a range of treatments that induce chromatin bridges during cell division. As shown in Fig. 1B concentrates in the centre of chromatin bridges between late telophase nuclei ( Fig. 1C and   S7D).

LEM-3 is a midbody-tethered protein
Given the LEM-3 staining pattern, we speculated that LEM-3 might localize at the midbody, which defines the site of cell abscission, the complete separation of two daughter cells at the end of cytokinesis. Using a strain carrying YFP-LEM-3 and mCherry-histone H2B fusions we found that YFP-LEM-3 localizes to the midbody region starting from telophase (Fig. 1D, white arrows; Movie S1). To establish if LEM-3 indeed colocalizes with the midbody, we compared the relative localization of LEM-3 and ZEN-4. ZEN-4 is a component of the centralspindlin complex, comprised of the ZEN-4/MKLP1 kinesin 6 motor protein and the CYK-4 Rho family GAP 20 , both essential for midbody formation and cytokinesis. We observed that LEM-3 colocalizes with ZEN-4 at the midbody ( Fig. 2A; Movie S2). Furthermore, midbody localisation of LEM-3 could not be observed upon RNAi depletion of CYK-4 or SPD-1, proteins essential for the integrity of the midbody, ( Fig. 2B and 2C), even in the presence of chromatin bridges induced by mcm-7 RNAi ( Fig. 2B and 2C). We also investigated if LEM-3 localization required cleavage furrow ingression. As expected, depletion of a contractile ring component NMY-2 inhibited furrow formation and ingression but didn't affect the central spindle formation (Fig.   S3A). We found that LEM-3 could still be detected at the midzone upon nmy-2 RNAi (Fig.   S3A). In summary, the LEM-3 nuclease associates with the midbody, and LEM-3 midbody location depends on the formation of the central spindle and the midbody.

LEM-3 is required for the resolution of chromatin bridges
We next sought to determine whether LEM-3 is required to resolve chromatin bridges that result from incomplete DNA replication, defective chromosome condensation, or unresolved recombination intermediates. We found that LEM-3 started to accumulate earlier at the midbody, and the intensity of LEM-3 foci was stronger in the presence of chromatin bridges generated by mcm-7 RNAi or the condensin I subunit capg-1 RNAi (Fig. 1D depletion no chromatin bridges formed in lem-3 mutants ( Figure S4A and S5C; 10/10 embryos). It thus appears that DNA bridges arising from incomplete DNA replication are efficiently processed in wild type, but remain unresolved in lem-3 mutants. Consistent with this view, we found that chromatin bridges were eventually resolved at the end of the first cell division in 33% of wild type embryos exposed to the 100% mcm-7 RNAi (n=9), while bridges persisted in all lem-3 mutant embryos treated with 100% mcm-7 RNAi (n=14) (Fig. 3B).
To further explore the functional significance of chromatin bridge resolution by LEM-3, we examined embryonic viability in either lem-3 mutants alone or in combination with non-lethal low-doses RNAi against mcm-7 or capg-1. Whereas 72% and 76% of wild type worms survived partial RNAi depletion with mcm-7 RNAi (2% bacteria expressing mcm-7 RNAi) or capg-1 RNAi (20% bacteria expressing capg-1 RNAi) respectively, the same RNAi treatment led to only 0.6% and 12.8% viable embryos in lem-3 mutants ( Fig. 3C and 3D). The hypersensitivity of lem-3 mutants to the partial inhibition of MCM-7 and CPAG-1 indicates that LEM-3 is important to process DNA intermediates that result from incomplete DNA replication or partial chromosome decondensation to ensure viability. We then investigated whether the same was true for DNA bridges that result from incomplete processing of DNA damage. As described above, lem-3; brc-1 double mutants show increased sensitivity to IR compared to the single mutants (Fig. S1A). This loss of viability in the double mutant is associated with extensive chromatin bridge formation, and lagging chromosomes and micronuclei became apparent, especially upon real-time imaging (Fig. S4, Movie S4). It thus appears that LEM-3 is important to process DNA bridges that result from the incomplete processing of DNA breaks when BRC-1 function is compromised.
Other endonucleases such as MUS-81 and SLX-1-SLX-4 have been shown to resolve recombination intermediates during metaphase (for review 7 ). To investigate the contribution of LEM-3 to recombination intermediates processing in relation to these other endonucleases, revealed that in 32% of lem-3; slx-4 embryos (n=19) cytokinesis failed, leading to binucleated cells and highly abnormal mitotic cell divisions ( Fig. S5B and S5D). In summary, our data suggest that LEM-3 processes recombination intermediates, as well as DNA structures resulting from replication stress or partial chromosome decondensation, and that such processing helps to maintain genome integrity.

The recruitment of LEM-3 to the midzone/midbody is important for chromatin bridge resolution
To better correlate the spatiotemporal behaviour of chromatin bridge resolution and the LEM-

Regulation of LEM-3 midbody localization and function by AIR-2/Aurora B kinase
The Aurora B kinase dependent NoCut checkpoint is able to detect chromatin bridges and delay abscission, allowing time for chromatin bridge resolution 4,8,19 . We found that the C.
Using Manders' overlay coefficient 77.4%±9.3% of LEM-3 protein co-localized with AIR-2 at the midzone/midbody (Fig. 4A). In addition, LEM-3 localization depends on the AIR-2/Aurora B kinase (Fig. 4B), indicating that AIR-2 regulates the midbody association and/or activity of LEM-3. We thus searched for AIR-2/Aurora B consensus phosphorylation sites (K/R; K/R; X0-2; S/T) in the LEM-3 protein sequence and identified two serines, Ser192 and Ser194, embedded in such a consensus sequence (Fig. 4C). The corresponding sites are phosphorylated by AIR-2/Aurora B in ZEN-4/MKLP-4, and the first serine also occurs in human LEM-3 orthologue Ankle1 23 . We thus mutated serines 192 and 194 to alanine by genome engineering ( Figure 5A). We observed comparable expression levels of GFP-LEM-3 S192AS194A and the wild type control (GFP-LEM-3, Fig. 5B). However, the lem-3 S192AS194A mutant showed an increased sensitivity to IR upon treatment of L4 stage larvae and late stage embryos ( Fig. 5C and 5D). In addition, combining the lem-3 S192AS194A mutations with slx-4 resulted in reduced viability compared to slx-4 mutant, consistent with increased levels of chromatin bridges observed in lem-3 S192AS194A; slx-4 double mutants To provide direct evidence that LEM-3 S192 and S194 are phosphorylated in vivo, we generated phospho-specific antibodies corresponding to a peptide containing both phosphorylated S192 and S194 (Fig. S7A). Using these LEM-3 phospho-specific antibodies we detected LEM-3 midbody-staining that overlapped with the localization of GFP-LEM-3 ( Fig. 6A). The same co-localization occurred on chromatin bridges induced by capg-1 and mcm-7 RNAi ( Fig. 6B and S7D). Importantly, no LEM-3 phosphorylation was detected on the midbody, when lem-3 S192AS194A and lem-3 (tm3468) mutants (the latter containing an inframe deletion that abrogates phosphorylation sites) were stained with the phospho-specific antibody ( Fig. S7B and S7C). Whereas GFP-LEM-3 is detected between nuclei from anaphase onwards in the presence of chromatin bridges induced by capg-1 RNAi, phosphorylated LEM-3 is only detected at the midbody during telophase (Fig. 6B). Therefore phosphorylation of LEM-3 might not be required for the initial recruitment but for congression of LEM-3 into the midbody. In summary, our data indicate that AIR-2 is required for robust localization of LEM-3, and LEM-3 phosphorylation may be important for stable association of LEM-3 with the midbody.

The GIY-YIG nuclease motif is essential for in vivo function of LEM-3
To determine whether the conserved GIY-YIG nuclease motif is required for LEM-3 function, we mutated the catalytic Y556 and G558 residues, which are predicted to be essential for nuclease activity 24,25 . Interestingly, LEM-3 localization at the midzone/midbody was not detectable in the Y556A G558A double mutant even after induction of chromatin bridges by capg-1 RNAi ( Fig. 5E; S10) suggesting that the mutations we introduced into the GIY-YIG motif might also compromise DNA binding and or midbody localization. lem-3 Y556A G558A has the same level of protein expression as wild type but is hypersensitive to IR (Fig. 5B, 5C and 5D). Moreover, combining the lem-3 Y556A G558A mutation with slx-4 led to 100% embryonic lethality and extensive chromatin bridge formation ( Fig. S5A and S5B). In addition, excessive and persistent chromatin bridge formation occurred upon capg-1 RNAi in lem-3 Y556A G558A embryos ( Fig. 5E and Movie S10). These results suggest that the GIY-YIG motif is essential for LEM-3 function in vivo.

Discussion
It has been known for a decade that the Aurora B mediated NoCut checkpoint plays a key role in delaying abscission, when chromatin is trapped at the midbody during cytokinesis 4,8 .
However, it remained enigmatic how chromatin bridge resolution occurs and how this process is connected with the abscission machinery and the NoCut checkpoint. In this study we report that the conserved endonuclease LEM-3 accumulates at the midbody and provide evidence that it resolves chromatin bridges at the final stage of cell division (Fig. 6C). LEM-3 midbody localization requires central spindle formation and is regulated by AIR-2/Aurora B kinase (Fig.   6C). Together with our evidence that LEM-3 does not act within the known major DNA repair pathways, this suggests that LEM-3 defines a new mechanism for maintaining genome integrity that acts just before the final stage of cytokinesis to resolve chromatin bridges.
In addition to its role in delaying abscission in the presence of chromatin bridges by inducing a stable intercellular canal at the midbody, our data indicate that at the same time Aurora B also promotes bridge resolution via recruitment of LEM-3 to the midbody. We found that the worm Aurora B kinase AIR-2 co-localizes with LEM-3 and is required for LEM-3 association with the midbody (Fig. 4A and 4B). Although we have no evidence for the direct interaction, LEM-3 has a conserved AIR-2/Aurora B phosphorylation site and can be phosphorylated in vivo, suggesting that LEM-3 may be a substrate of AIR-2/Aurora B kinase. In addition, the occurrence of phosphorylated LEM-3 during late cytokinesis is consistent with the reported high level of AIR-2 kinase activity at this stage. It is noteworthy that the lem-3 S192A S194A phosphosite mutant is not as sensitive to IR as the lem-3 null mutant. This might indicate that mutation of these two serines impairs but does not completely abolish LEM-3 function (Fig.   5C, 5D and 5E). This may be due to the residual midbody location of LEM-3 in this mutant ( Fig. 5E). Our finding that midzone/midbody formation plays a key role in chromatin bridge resolution further supports the importance of LEM-3 midbody localization. The C. elegans checkpoint preferentially works after the second embryonic cell cycle 19 , rendering chromatinbridge processing by LEM-3 more important in the first cell cycle. Indeed, the faster separation of nuclei we observed in lem-3 mutants may suggest that LEM-3 also has a direct role in the NoCut checkpoint. Irrespectively, while we provide strong evidence that LEM-3 acts on the midbody during cytokinesis, we cannot rule out that LEM-3 might also have a role in earlier stages of the cell cycle.
We found that LEM-3 is able to bind to and resolve chromatin bridges caused by various perturbations, including incomplete DNA replication, unresolved recombination intermediates, and compromised chromosome condensation (Fig 1B, 3A, S2A and S5), indicating that LEM-3 can process a wide range of DNA substrates with distinct structures. In organisms with large genome, unreplicated DNA is frequently present in normal cells at the end of G2 phase 26 . The post-replicative resolution of unreplicated DNA depends on creation of ultrafine anaphase bridges during mitosis 26 . Ultrafine DNA bridges, which cannot be stained by conventional DNA dyes are different from chromatin bridges 27 . It is possible that the ultrafine DNA bridges exist in early C. elegans embryos where rapid cell divisions occur. Ultra fine bridges in mammalian cells are detected by being coated with the BLM or the PICH helicase [28][29][30] . We did not observe bridges coated with the C. elegans BLM ortholog HIM-6, and PICH has not been found encoded in the worm genome (our unpublished data). Irrespective, in order to segregate unreplicated segments of DNA, the unreplicated double stranded DNA has to be disentangled to form single stranded DNA, which can be repaired in the next cell cycle 26 . Therefore, the accumulation of LEM-3 at the midbody during a late stage of cell division may also contribute to the resolution of the ultrafine DNA bridges caused by unreplicated DNA regions in unperturbed cells.
In human cells, the cytoplasmic nuclease TREX1 was recently found to contribute to the cutting of chromatin bridges that originate from telomere fusion 31 . TREX1 binds to the chromatin bridges after anaphase, and its nuclease activity is important for their timely cleavage. Unlike LEM-3, whose localization becomes more restricted at the midbody as cytokinesis proceeds, TREX1 is present throughout chromatin bridges and generate extensive ssDNA. Moreover, TREX1 is an exonuclease that requires nicked DNA substrates, indicating that TREX1 is not solely responsible for chromatin bridge resolution, or else can only act on nicked DNA substrates. Furthermore, in contrast to LEM-3, which processes chromatin bridges at the end of cell division to maintain genome stability and allow for increased survival, this is not clear for TREX1, which might only get access to bridges when the nuclear envelope is perturbed. In addition, the cutting of telomere-telomere bridges by TREX1 leads to pathological 'McClintock-like' breakage-fusion-bridge cycles often linked with a large number of randomly assembled chromosome fragments, a phenomenon termed Chromothripsis [31][32][33] . In contrast, our data suggest that LEM-3 is able to process a variety of DNA bridges to promote viability ( Fig. 3C and 3D). Thus LEM-3 is part of a 'last chance saloon' mechanism to maintain genome integrity.
It will be very interesting in future studies to explore further the role of the Ankle1 nuclease 10 , which is the ortholog of C. elegans LEM-3 in human and mouse. The localization of endogenous Ankle1 has neither been reported, nor its potential association with chromatin bridges explored, though over-expressed Ankle1 protein was found to be cytoplasmic 10 . Mice lacking Ankle1 do not have an overt phenotype and extracted cells are not hypersensitive to DNA damaging agents 11 , but this is also true of mice lacking the Gen1 or Mus81 nucleases, indicating a high level of redundancy between DNA nucleases that preserve genome integrity in mammalian systems 34 . Interestingly, increased susceptibility to breast and ovarian cancer in the general population and in carriers of BRCA1 and BRCA2 mutations was mapped down to 13 polymorphisms at 19p13.1 mapping to non-coding regions in the human LEM-3 homolog Ankle1 and a second gene ABHD8, encoding for a alpha beta hydrolase domain [13][14][15] . The effect of these polymorphisms on Ankle1 function remains to be explored, but our finding that lem-3; brc-1 double mutants show elevated sensitivity to DNA damage ( Fig. S1A and S4B) is consistent with Ankle1 being the most likely gene in 19p13.1 to affect breast cancer disposition. Indeed, another study found that coding mutations in Ankle1 were associated with increased risk of breast cancer 35 . It will be important in the future to explore whether Ankle1 and BRCA1 cooperatively promote genome stability in human cells, as predicted by the behaviour of their C. elegans equivalents.

Experimental Model and Subject Details
C. elegans strains were maintained at 20°C on nematode growth medium (NGM) plates seeded with OP50 bacteria. Transgenic lines were generated by injecting plasmid DNA/ssDNA directly into the hermaphrodite gonad. Integrated transgenic strains were outcrossed at least six times before use.

Embryonic viability assay
L4 or young adult worms were treated with various DNA damaging agents as previously described 36 . After 24 hours of recovery, worms were transferred onto new plates and allowed to lay eggs for 6-8 hours. Eggs were quantified. Unhatched eggs were counted 24 hours later and the percentage of progeny viability was calculated.

Embryo irradiation assay
The embryo irradiation assay was performed as described previously 17,36 . Experiments were done in triplicate. 10 young adult worms of each strain were transferred onto seeded plates.
Worms were allowed to lay eggs for 2 h and were removed from the plate afterwards. The embryos were incubated at 20°C for 3 h before treatment with different doses of IR (0 Gy, 30 Gy, 60 Gy, 90 Gy). The eggs were counted, and the plates were incubated at 20°C for 24 h.
Unhatched eggs were counted the next day and plates were incubated for another 24 h. The delayed development phenotype was scored 48 hours after IR by calculating the percentage of worms that reached the L4 stage. OMX microscopy, established protocols were followed 38 . To record mitotic divisions, we followed previously described imaging procedures 18 . Embryos were dissected in M9 (3 g/l KH2PO4, 6 g/l Na2HPO4, 5 g/l NaCl, 1 mM MgSO4) and mounted on 2% agarose pads.

Image acquisition
Images were captured every 10 s at 23-24°C using a spinning-disk confocal microscope (IX81; Olympus) with spinning-disk head (CSU-X1; Yokogawa Electric Corporation) and MetaMorph software (Molecular Devices). Image analysis and video processing were performed using ImageJ software (National Institutes of Health).
Exact details are available upon request. For generated point mutations in the endogenous lem-3 locus, we performed CRISPR-Cas9 following procedures described by the Seydoux lab 40

Western blotting
Wild type, GFP-lem-3, GFP-lem-3 S192AS194A and GFP-lem-3 Y556AY558A adult worms were collected and bleached until they were completely lysed. After centrifugation for 1 min at 2200 rpm the pellet was resuspended in 200 µL M9 worm buffer. The suspension was transferred into safety capped 1.5 mL reaction tubes. 200 µL of 40% TCA and 300 µL of glass beads were added and the worm eggs were disrupted by bead beating for 2 min. The suspension was then transferred into a 1.5 mL reaction tube. The beads were washed with 5% TCA which was added to the suspension afterwards. Samples were spun down for 10 min at 3000 rpm, the supernatant was removed and the pellet was resuspended in 1 mL ice-cold acetone. After another centrifugation for 2 min at 11,000 rpm the acetone was discarded and the pellet was left to dry at room temperature. Finally the pellet was resuspended in 100 µL Laemmli buffer (3x). 10 µL of each sample was loaded onto a 4%-12% NuPage precast gel.