Structural Maintenance of Chromosome (SMC) Proteins Link Microtubule Stability to Genome Integrity*

Background: SMC proteins organize chromosome architecture, whereas microtubules form the structural framework for chromosome movement in mitosis. Results: SMC proteins bind to and stabilize microtubules to promote proper mitotic spindle formation. Conclusion: SMC proteins can provide a functional link between microtubules and DNA to ensure faithful chromosome segregation. Significance: SMC-microtubule interactions are essential to establish a robust system for maintaining genome integrity. Structural maintenance of chromosome (SMC) proteins are key organizers of chromosome architecture and are essential for genome integrity. They act by binding to chromatin and connecting distinct parts of chromosomes together. Interestingly, their potential role in providing connections between chromatin and the mitotic spindle has not been explored. Here, we show that yeast SMC proteins bind directly to microtubules and can provide a functional link between microtubules and DNA. We mapped the microtubule-binding region of Smc5 and generated a mutant with impaired microtubule binding activity. This mutant is viable in yeast but exhibited a cold-specific conditional lethality associated with mitotic arrest, aberrant spindle structures, and chromosome segregation defects. In an in vitro reconstitution assay, this Smc5 mutant also showed a compromised ability to protect microtubules from cold-induced depolymerization. Collectively, these findings demonstrate that SMC proteins can bind to and stabilize microtubules and that SMC-microtubule interactions are essential to establish a robust system to maintain genome integrity.

Structural maintenance of chromosome (SMC) proteins are key organizers of chromosome architecture and are essential for genome integrity. They act by binding to chromatin and connecting distinct parts of chromosomes together. Interestingly, their potential role in providing connections between chromatin and the mitotic spindle has not been explored. Here, we show that yeast SMC proteins bind directly to microtubules and can provide a functional link between microtubules and DNA. We mapped the microtubule-binding region of Smc5 and generated a mutant with impaired microtubule binding activity. This mutant is viable in yeast but exhibited a cold-specific conditional lethality associated with mitotic arrest, aberrant spindle structures, and chromosome segregation defects. In an in vitro reconstitution assay, this Smc5 mutant also showed a compromised ability to protect microtubules from cold-induced depolymerization. Collectively, these findings demonstrate that SMC proteins can bind to and stabilize microtubules and that SMC-microtubule interactions are essential to establish a robust system to maintain genome integrity.
During cell division, genomic DNA must be partitioned to daughter cells in a highly regulated fashion to ensure the maintenance of genome integrity from each generation to the next. High fidelity segregation of eukaryotic genomes requires pre-cise coordination between the multiple components of the cellular segregation machinery. For instance, replicated sister chromatids must be connected to each other and reconfigured into highly condensed chromosomes to allow effective segregation in anaphase (1,2). This reorganization of chromatin depends in large part on a family of proteins known as the structural maintenance of chromosome (SMC) 6 proteins (2)(3)(4). SMC proteins assemble into large multisubunit complexes (cohesin, condensin, and the Smc5-Smc6 complex) that bind to chromatin in mitosis, where they promote the assembly of functional chromosomes. Once mitotic chromosomes are assembled, they become competent for attachment to a microtubule-based structure, the mitotic spindle, and for subsequent segregation in anaphase.
The microtubule-based cytoskeleton also undergoes dramatic changes and intricate regulation during the cell cycle. Microtubules switch from providing a structural role for the maintenance of cell shape and for intracellular transport in interphase to building the underlying framework for chromosome segregation during mitosis. Mitotic chromosome formation and spindle assembly are often considered to be two precisely coordinated but separate events because proteins that are involved in chromatin organization and remodeling are often distinct from those that regulate cytoskeletal structures. Prime examples include histones, which are not linked to any polymer structures other than chromatin. Similarly, many known microtubule-associated proteins, including the XMAP215 family of TOG domain-containing proteins and the EB family of proteins, are not known to interact with chromatin (5,6), and yet, during cell division, the mitotic spindle microtubules need to make contact with the chromosomes in order to move and segregate them to opposite poles of the cell. Therefore, proteins linking chromosomes to the spindle apparatus are of great interest because they can provide fundamental insights into the mechanisms responsible for high fidelity genome segregation in eukaryotes. Microtubule-chromosome connections are thought to be mediated in part through the kinetochore (7-9), a large multi-protein complex assembled on the centromeric region of chromosomes. Microtubules can also make direct contact with chromosome arms via proteins endowed with dual affinity for chromatin and microtubules (e.g. motor proteins of the chromokinesin family (10 -12)). It is currently unclear how many different protein families can mediate direct interaction between chromatin and microtubules and what their specific contributions are to high fidelity chromosome segregation.
There have been a number of proteomic studies on microtubule-interacting proteins reported in the literature (13)(14)(15)(16). Interestingly, some chromatin-binding proteins, most notably SMC family members, were consistently identified as microtubule-interacting proteins in such studies (13,15). The fact that these microtubule interactomes are devoid of other abundant chromatin-associated proteins, such as histones, suggests that the SMC family members identified in these studies are not simply contaminants or artifacts. This raises the possibility that SMC proteins directly bind to microtubules, an unexpected property for this family of chromatin-binding proteins. However, formal proof of this possibility is still lacking. Given the biological significance of proteins linking microtubules and chromatin, we set out to determine whether SMC proteins interact directly with microtubules and to understand the potential roles and implications of these interactions in vivo.

Protein Purification
Full-length Smc5-STH was expressed and purified in yeast, as described previously (17). Smc5 fragments were expressed in bacteria from pET-series vectors (Novagen) and purified by nickel chelate and/or Strep-Tactin chromatography (Qiagen), according to the manufacturers' suggested conditions. The condensin complex containing Brn1, Smc2, Smc4, Ycg1, and Ycs4 subunits was purified from budding yeast by sequential nickel chelate, Strep-Tactin, and gel exclusion chromatography, as described previously (18). All proteins were purified to Ն95% purity, as judged by Coomassie Blue staining. Purification of bovine tubulin and the labeling of tubulin with X-Rhodamine dye were performed as described (19).

Microtubule Binding Experiments
Co-sedimentation Assay-The co-sedimentation assay was performed as described previously with some modifications (20). Briefly, microtubules were prepared by polymerizing 25-40 M tubulin in 1ϫ BRB80 (80 mM PIPES, 1 mM EGTA, 1 mM MgCl 2 ), 1 mM GTP, 1 mM DTT in the presence of 10% DMSO at 37°C for 30 min and then diluted to 10 or 20 M working stocks in 1ϫ BRB80 with 20 M paclitaxel. Binding reactions were done by mixing SMC proteins with microtubules in 1ϫ BRB80 supplemented with the indicated salt concentrations (75-125 mM NaCl), 1 mM DTT, 0.01% Tween 20, and 20 M paclitaxel. After a 20-min incubation, mixtures were spun at 45,000 -60,000 rpm in a Sorvall S120AT3 rotor at 25°C for 5 min. Samples from supernatant and pellet fractions were resolved by SDS-PAGE. Gels were stained by Coomassie Blue and scanned with a digital scanner. Band intensities were quantified using ImageJ (NIH). K d measurements were done as described previously (20).
Microscopy-based Microtubule Binding Assays-Microscopy-based microtubule binding assays were performed as described previously (20). For detection of the microtubule-Smc5-DNA ternary complex, samples of Smc5-bound X-Rhodamine-labeled microtubules were first generated by mixing of the two components at a molar ratio of ϳ1:4 (Smc5/tubulin dimer) and then attached them onto a tubulin antibody-coated (rat ␣-tubulin antibody from Serotec, at 10 ng/ml) coverslip in a flow chamber. Fluorescently labeled DNA (6-carboxyfluorescein (FAM)-labeled 60-bp single-stranded DNA, Integrated DNA Technologies) was then introduced into the chamber. After a 5-min incubation, unbound DNA was flushed out, and images of the bound DNA and microtubules were visualized and recorded on a Zeiss Axio-Imager Z1 microscope equipped with a 1.4 numerical aperture Plan-APOCHROMAT 63ϫ objective using the Zeiss AxioVision release 4.8 software. The TIFF files of the acquired images were then processed by Meta-Morph (Molecular Devices, LLC).

Image Analysis for Microtubule-Smc5-DNA Interactions
The quantitative analysis of the microtubules and the associated FAM-labeled DNA was done using MetaMorph (Molecular Devices). To identify microtubules, the images of the X-Rhodamine-labeled microtubules for the no Smc5 control were first thresholded to a signal intensity that would allow the detection of individual microtubules in the control image while minimizing background noise. The objects were identified using the "integrated morphometry analysis" functionality with exclusion of objects that touched the image edge and those with an area smaller than 50 square pixels (non-microtubule signals or X-Rhodamine aggregates). The same thresholding and exclusion filtering were applied to the acquired images with Smc5-WT and Smc5-3KE. The average area and average signal intensity of all of the objects were obtained automatically using the integrated morphometry analysis. A conservative cut-off of area occupied by microtubules was set to distinguish single microtubules from bundles. Object identification was verified by visual inspection. To analyze the extent of the FAM-oligonucleotide binding to microtubules, regions were created around the identified objects in the microtubule images and were transferred onto the corresponding images containing the FAM signal. This allows the determination of the average signal intensity within the microtubule delineated areas. For all calculations, the average background signal intensities (from unoccupied areas) were subtracted from the average signal intensity of microtubules and FAM, respectively. Using this analytical method, the average area occupied by single microtubules in the control, the average number of microtubules within bundles and the FAM intensity per microtubule were calculated.

Determination of Candidate Microtubule Binding Sites
To prepare biotin-labeled Smc5 protein that has impaired microtubule binding ability, purified Smc5 proteins were chemically modified in the presence of 0.1-0.5 mM NHS-LCbiotin (Pierce/Thermo Fisher). Biotinylation reactions were performed at room temperature for 5 min and quenched by the addition of potassium glutamate. Modification was verified by streptavidin-HRP Western blot (streptavidin-HRP from Pierce/Thermo Fisher, 1:2000 dilution). Biotinylated Smc5 proteins were allowed to bind to microtubules and subjected to a sedimentation assay. This allows the separation of the fraction that associated with microtubules (pellet) and the fraction that remains in the supernatant (enriched for protein with biotinylated sites that interfere with binding). Optimal biotinylation and microtubule binding conditions were determined experimentally. Smc5 proteins in the supernatant fraction and the pellet fraction were analyzed by mass spectrometry. Intensities of the trypsin-digested peptides were normalized and compared between the two fractions. Biased enrichment of any biotinylated peptides in the supernatant fraction indicated lysine sites that were potentially essential for microtubule binding.

Cold-induced Microtubule Depolymerization Assay
The assay was performed as previously described (21). Briefly, stock microtubules were diluted 10-fold in 1ϫ BRB80 with 2 M paclitaxel and 75 mM NaCl either in the absence or presence of the Smc5 proteins. The mixtures were then chilled to 2°C in a PCR machine and incubated for 30 min to induce microtubule depolymerization while the control sample was placed at room temperature. The samples were then subjected to ultracentrifugation to separate the depolymerized tubulin dimers from microtubules and processed as described above in the co-sedimentation assay.

Yeast Genetics and Growth Conditions
Yeast culture and genetic modifications were performed according to St-Pierre et al. (18). The genotypes of yeast strains used in this study are described in Table 1. smc5 mutants defective in microtubule-binding residues were created by site-directed mutagenesis (QuikChange, Stratagene) and integrated into the yeast genome, as described previously (17). The presence of the appropriate mutations in the yeast genome was confirmed by sequencing the SMC5 locus. Cell proliferation at various temperatures and/or in the presence of DNA-damaging agents was determined by spotting 5-fold dilution series of yeast cultures on solid medium containing the appropriate supplements and/or drugs. Cells were grown until individual colonies became visible on the surface of the solid media (2-3 days between 23 and 37°C; 4 -5 days at 18°C). A previously characterized temperature-sensitive mutant in the SMC5 gene, smc5-6, was included in some experiments as a control for the effect associated with losing Smc5-Smc6 complex function (22).

Yeast Cell Cycle Experiments
Cell cultures were arrested in the G 1 phase of the cell cycle using ␣-factor (50 ng/ml for 180 min) and synchronously released from this arrest at 18°C. Samples were taken at intervals and processed for microscopic characterization of yeast bud, spindle, and nuclear morphology, as described previously (23). Yeast images were acquired on a Leica DM5500B microscope equipped with an EM-CCD camera (Hamamatsu). Volocity software (PerkinElmer Life Sciences) was used to acquire images. Brightness and contrast adjustments were applied to primary micrographs using Photoshop software (Adobe). During time course experiments, yeast culture samples were also processed for determination of DNA content by flow cytometry using published procedures (23).

RESULTS
Smc5 Binds Directly to Microtubules in Vitro-To test the possibility that SMC proteins bind to microtubules directly, we first tested the purified condensin complex in a microtubule co-sedimentation assay and found it to be associated with the polymers. Indeed, Fig. 1A shows that the five subunits of the yeast condensin complex (i.e. Brn1, Smc2, Smc4, Ycg1, and Ycs4) (18) are enriched in the pellet fraction containing micro- tubules after sedimentation. We next sought to determine whether this microtubule-binding behavior is also typical of other SMC complexes. Specifically, we focused on Smc5 because this protein can be purified in recombinant form (17). Smc5 is a member of the Smc5-Smc6 complex related structurally to the condensin complex but completely distinct in subunit composition. The fact that it can be purified in monomeric form enabled us to determine whether SMC subunits are directly responsible for the interaction with microtubules within their own complexes. We found that Smc5 also associated with microtubules in the co-sedimentation assay, indicating a direct interaction between these proteins (Fig. 1B). Judging by the observation that more than half of the SMC proteins associated with microtubules (at 0.5 M tubulin dimer concentration) in these experiments, we concluded that SMC-microtubule interaction was of high affinity, with a dissociation constant (k d ) of about or below 0.5 M. It should be noted that the ionic strength of the buffer condition used in these experiments (80 mM PIPES with ϳ125 mM NaCl) was much higher than those used traditionally (ranging typically from 12 to 80 mM PIPES or 20 mM sodium phosphate with 0 -50 mM KCl or NaCl) (24 -26). Although the condensin complex is an ATPase (2), its association with microtubules did not appear to be affected by the presence of different nucleotides (ADP, ATP, or the nonhydrolyzable ATP analog AMP-PNP). The condensin complex did, however, appear to bind more to microtubules in the absence of nucleotide (Fig. 1C). This is distinct from the nucleotide dependence of many prototypical kinesin motor proteins, where ATP hydrolysis or ADP releases motors from microtubules (24,27). Interestingly, although Smc5 is capable of hydrolyzing ATP as a monomer (17), possible differences in the levels of ATPase activity present in the purified Smc5 and condensin fractions may explain their differential sensitivity to nucleotides in microtubule-binding assays (Fig. 1C).
To simplify our investigation, we decided to follow the SMC component of the Smc5-Smc6 complex and chose purified Smc5 for subsequent experiments (17). We reconfirmed the Smc5-microtubule interaction using a microscopy-based flow chamber assay (Fig. 1D). In this assay, Smc5 protein was nonspecifically absorbed onto a glass coverslide in a flow chamber. Fluorescently labeled microtubules were then infused into the chamber to allow binding to the surface-immobilized Smc5. As observed in Fig. 1E, microtubules were efficiently captured only in the chamber that had been coated with Smc5 protein, again indicating a direct interaction.
Mapping and Assessment of the Smc5-Microtubule Interaction-Having demonstrated the ability of Smc5 to bind directly to microtubules, we next asked what the functional significance of this interaction would be. One way to address this question is to identify and characterize a mutant with impaired microtubule binding ability. However, without a crystal structure or a recognizable microtubule-binding motif in Smc5, locating the interaction interface in a ϳ125-kDa protein is not a trivial task. To narrow our search, we decided to map the minimal binding domain using truncated forms of Smc5 containing only the ATPase heads, or the hinge domain with medium (HM) or long (HL) coiled-coil extensions ( Fig. 2A). By a co-sedimentation assay, we found that the hinge and the  Note that due to the large differences in the molecular mass of tubulin and condensin subunits, it is difficult to resolve all individual components of this complex in a single gel while simultaneously visualizing tubulin subunits (hence the appearance of only three major bands at the position of the condensin complex in the gel (18)). The weak band above the major Ycg1-Smc2 band corresponds to phosphorylated Ycg1, as reported previously (18).   (Fig. 2, B and C). This finding is consistent with our finding that Smc5-microtubule interaction is not sensitive to the nucleotide-binding states of Smc5 (Fig. 1C). Next we determined whether Smc5, like many microtubule-associated proteins, interacts with microtubules via the C-terminal negatively charged tails of tubulins known as the E-hooks (28 -31). To achieve this, we used the protease subtilisin to cleave off the E-hooks from microtubules and found that this treatment did indeed drastically reduce Smc5 binding (Fig. 2D). This shows that Smc5 interacts with microtubules largely by electrostatic interactions via the C-terminal tails of tubulins.
Identification and Characterization of Microtubule-binding Mutants of Smc5-Although we had narrowed down the domains that contain the bulk of the microtubule binding activity of Smc5, we still needed to pinpoint the key regions or amino acid residues that mediate the interaction. To achieve this, we took advantage of the observation from the subtilisin experiment that negative charges played a role in the microtubule-Smc5 interaction to develop a strategy to obtain a microtubulebinding mutant (Fig. 3A). We hypothesized that positively charged residues on Smc5 are important to interact with the E-hooks of tubulins, so we chemically modified lysine residues in Smc5 with biotin under non-saturating conditions. Our rationale was that the labeling of lysine residues essential for microtubule interaction would impair binding to the polymers, whereas that of unrelated lysine residues would not. By combining this approach with the microtubule co-sedimentation assay, it is possible to obtain a biased distribution of differentially labeled Smc5 protein molecules. Smc5 with those lysine residues "essential" for microtubule-binding biotinylated will be enriched in the supernatant fraction (i.e. non-microtubule binding), whereas those with lysines not relevant to microtubule-binding biotinylated will remain associated with the microtubule pellet, behaving similarly to the non-modified protein (Fig. 3B). Subsequent comparative mass spectrometry analysis of the supernatant and the pellet fractions will then determine which biotinylated lysines are specifically enriched in the individual fractions.
Using the strategy described above, we identified three residues (Lys-624, Lys-631, and Lys-667) in the microtubule-binding region of Smc5 (i.e. Smc5-HM) that were consistently enriched in the supernatant fraction in the microtubule co-sedimentation assay (n Ͼ 3). Using site-directed mutagenesis, we generated a lysine to glutamic acid mutant (called Smc5-3KE) to determine the effect of these mutations on microtubule binding. We purified Smc5-3KE protein from yeast and tested the mutant in a microtubule co-sedimentation assay. We found that, in comparison with its wild-type counterpart, Smc5-3KE had an ϳ2-fold decrease in binding affinity for microtubules DNA Binding Activity of Microtubule-binding Mutants of Smc5-Because Smc5 is a known DNA-binding protein (17), we wished to determine whether the 3KE mutations affect the DNA-binding affinity of this protein. The ability of purified wild-type and 3KE mutant forms of Smc5 to bind singlestranded DNA (ssDNA) was determined using an electrophoretic mobility shift assay (EMSA). As shown previously (17), we find that wild-type Smc5 is capable of fully binding/shifting ssDNA in the well of the gel at a protein/DNA ratio of 25-fold molar excess (Fig. 4A). The Smc5-3KE mutant could also fully shift the ssDNA to the well of the gel although at a slightly higher protein/DNA molar ratio (Fig. 4B). We thus concluded from these experiments that the Smc5-3KE mutant largely retains its DNA binding activity, although the mutations may have a slight effect on this activity. It should be noted that the DNA binding activity of SMC complexes is contributed by multiple independent subunits of these complexes (32)(33)(34) and that a modest loss in the activity of a single component is likely to be compensated by the DNA binding activity of other subunits.
Smc5 Can Bind to DNA and Microtubules Simultaneously-To determine whether Smc5 can bind to both DNA and microtubules simultaneously, we developed a microscopy-based assay to visualize ternary complex formation. In this assay, sur- A, schematic depicting the method used to identify surface-exposed lysine residues on Smc5 that are essential (red ϩ) or non-essential (blue ϩ) for microtubule binding. Upon chemical labeling by biotin, modified lysines essential for binding will be enriched in the supernatant fraction. Relative abundances of tryptic peptides containing the modified lysines in the supernatant and pellet fractions were quantified by mass spectrometric analysis. B, Coomassie Blue-stained gel and streptavidin-HRP blot of the supernatant and pellet fractions of mock-labeled or biotin-labeled Smc5-HM proteins from the microtubule co-sedimentation assay. The enrichment of labeled Smc5-HM in the supernatant indicates the perturbation of microtubule binding upon lysine modifications. C, three lysine residues in the hinge region of Smc5 identified by the method illustrated above in A and B were mutated into glutamic acid to generate the microtubule-binding mutant called Smc5-3KE. Coomassie Blue-stained microtubule titration gels of full-length Smc5-WT (wild-type) or Smc5-3KE from the co-sedimentation assay are shown. D, fractions of microtubule-bound Smc5-WT (red) and Smc5-3KE from four independent microtubule co-sedimentation experiments were plotted against microtubule concentrations and fitted to a hyperbola to determine their dissociation constants (K d ).

JOURNAL OF BIOLOGICAL CHEMISTRY 27423
face-anchored X-Rhodamine-labeled microtubules precoated with Smc5 were allowed to interact with fluorescently labeled (FAM-labeled) DNA in solution (Fig. 4C). Using this method, we observed the capturing of fluorescently labeled DNA by Smc5-coated but not naked microtubules, indicating the formation of microtubule-Smc5-DNA ternary complex (Fig. 4D).
It is worth noting that this capturing of FAM-labeled DNA was much reduced on Smc5-3KE-coated microtubules, as evidenced by the weak FAM signal that was only barely above that of the control condition. In addition, we found that microtubules were more efficiently bundled by Smc5 than by Smc5-3KE. Based on our quantification (Fig. 4E), Smc5-induced bundles represented ϳ90% of the input microtubules (versus 72% with Smc5-3KE), and each bundle consisted of ϳ17 microtubules on average (versus ϳ12 microtubules with Smc5-3KE). In addition, the average FAM intensity on Smc5-induced microtubules is substantially higher than that on Smc5-3KE bundles or control (Fig. 4F). Together, our results show that Smc5 can interact with both microtubules and DNA simultaneously, and 3KE mutations compromise the formation of the ternary complex.

Smc5 Interacts with Microtubules in Dividing Yeast
Cells-Having demonstrated the interaction between Smc5 and microtubules in vitro, we next sought to determine whether this interaction occurs in vivo. To test this notion, we performed a proximity ligation assay (PLA), an established method to detect  protein-protein interactions based on their proximity within 40 nm in distance (20,35,36). For this, we used antibodies against tubulin and the StreptagII-fused Smc5. Their close proximity was detected and amplified by the secondary antibody-conjugated fluorescently labeled Duolink probes (Fig. 5A). Specifically, we observed a significant increase in the number of cells showing a PLA-positive spot (indicating an in vivo Smc5-microtubule interaction) in the SMC5-STH yeast strains, as compared with that in the no-tag control (SMC5; Fig. 5, B and C). In addition, there was a specific enrichment of PLA signals in the nucleus and on microtubules in the StreptagII-fused Smc5 strain (Fig. 5D). Together, these results suggest a direct Smc5microtubule interaction in dividing yeast cells. Functional Significance of Smc5-Microtubule Interaction-To assess the physiological consequence of impaired microtubule binding on Smc5 functions, we integrated the smc5-3KE mutant at the endogenous SMC5 locus in yeast. We found that the smc5-3KE mutant could substitute endogenous Smc5 function for viability at 30 and 37°C. However, the mutant strain exhibited compromised growth kinetics at lower temperatures (Fig. 6A), consistent with a conditional cold-sensitive (cs) phenotype. The absence of detectable phenotype in diploid cells expressing smc5-3KE and wild-type SMC5 indicates that the mutant allele is recessive for cs behavior (Fig. 6A, bottom panels). smc5-3KE is the first identified mutant allele in any members of the SMC family that shows a conditional cs phenotype in budding yeast. Cold-sensitive phenotypes are relatively rare in essential proteins in eukaryotes and typically associate with defects in structural components of the cell (37)(38)(39). Consistent with this, tubulin mutants represent one of the largest classes of essential proteins for which cs mutations have been isolated (e.g. see the tub2-401 allele in Fig. 6A (38)). Microtubule mutants are known to be sensitive to lower temperatures at least in part because this condition can lead to microtubule depolymerization (40 -42). This, together with the shared cs phenotypes of tubulin and smc5-3KE mutants, led us to wonder whether Smc5 acts by stabilizing microtubule polymers. To test this idea, we developed an in vitro assay to reconstitute coldinduced depolymerization. In this assay, paclitaxel-stabilized microtubules were diluted to low concentration and subjected to low temperature (i.e. 2°C) to induce depolymerization. Compared with the identical sample at room temperature, the cold-treated sample had substantially more depolymerization, evident by the presence of more tubulins in the supernatant fraction in the sedimentation assay. Remarkably, substoichiometric ratio of Smc5 effectively stabilized microtubules at 2°C (Fig. 6, B and C). In contrast, Smc5-3KE had a significantly reduced ability to protect microtubules from cold-induced depolymerization. This result led us to hypothesize that the cs phenotype and growth arrest we observed in smc5-3KE mutant yeast were due to microtubule instability and activation of the spindle assembly checkpoint. To test this in vivo, we investigated whether double mutants defective in smc5-3KE and microtubule components (38) or spindle assembly checkpoint activity (43,44) show synthetic defects relative to single mutants. Remarkably, combining tub2-401 or mad1⌬ with the smc5-3KE allele in yeast resulted in a significant decrease in the proliferative properties of the double mutant, relative to the single mutants, at lower temperatures (Fig. 6, D and E). Importantly, the microtubule-related phenotype of the smc5 mutant was not the consequence of mutation-induced disassembly of the Smc5-Smc6 complex or abnormal abundance of the Smc5-3KE protein in cells (Fig. 7). Together, these data suggest that Smc5 plays a direct role in stabilizing microtubules in yeast.
We next wondered whether loss of microtubule binding affected the known functions of the Smc5-Smc6 complex in the DNA damage response (3). To test this notion, we monitored the cellular proliferation of smc5-3KE mutants in the presence of various DNA-damaging agents, including methyl methanesulfonate (MMS), 4-nitroquinoline 1-oxide (4NQO), and hydroxyurea (HU). Fig. 6F shows that loss of microtubule binding ability significantly reduced the capacity of the smc5-3KE mutant to grow in the presence of all three DNA-damaging agents tested. As observed previously for the cs phenotype, the DNA damage sensitivity of the smc5-3KE mutant was comple-  OCTOBER 3, 2014 • VOLUME 289 • NUMBER 40

SMCs Link Microtubules to Genome Integrity
mented by a single wild-type copy of SMC5, indicating that the mutant allele is recessive for its sensitivity to DNA-damaging agents (data not shown). Together, these observations indicate that Smc5 depends at least in part on its interaction with microtubules and/or on a normal microtubule network (see below) to maintain genome integrity in cells exposed to DNA damage.
If the binding of Smc5 to microtubule affects the stability and/or functionality of the microtubule cytoskeleton in vivo, one would expect that the smc5-3KE mutant should have mitotic phenotypes resembling those of tub2 mutants. To test this hypothesis, we performed a cell cycle experiment in which smc5-3KE mutants and wild-type yeasts where synchronized in G 1 at 30°C using ␣-factor and then released from this arrest  Smc5 stabilizes microtubules in vivo and in vitro. A, the smc5-3KE mutant exhibits a recessive cold-sensitive growth defect in yeast. 5-Fold dilution series of wild-type haploid cells, smc5-6 mutant (22), tubulin mutant tub2-401 (38), wild-type SMC5, or smc5-3KE, were spotted on solid medium to evaluate growth at the indicated temperatures of 18, 23, 30, and 37°C. Bottom panels, proliferation of diploid yeast heterozygous for SMC5-STH or smc5-3KE-STH on solid medium at various temperatures. The absence of cold-sensitive growth defect in the diploids indicates that the smc5-3KE allele is recessive. B, Smc5 protects microtubules from cold-induced depolymerization. After incubation at the indicated temperature (2°C or room temperature (RT)), reaction mixtures containing buffer control (Ϫ), Smc5-WT, or Smc5-3KE were subjected to ultracentrifugation to separate the dissociated tubulin dimers and the remaining microtubule polymers into supernatant (S) and pellet (P) fractions, respectively. Coomassie Blue-stained gels of the samples are shown. The relative quantity of the Smc5 proteins in arbitrary units (a.u.) in each sample is shown below the corresponding lanes on the gel. C, quantification of the percentage of protection provided by Smc5 proteins (average of three independent experiments; error bars, S.D.). Percentage protection is defined as percentage of excess microtubules retained in the pellet relative to the control at 2°C. D and E, genetic interactions between smc5-3KE and the tubulin-defective mutant tub2-401 (D) or the spindle assembly checkpoint mutant mad1⌬ (E). Growth phenotypes of the indicated yeast strains were evaluated as described in A. F, proliferation of smc5 mutants and wild-type yeast strains on solid medium containing methyl methanesulfonate (MMS) at 0.005%, hydroxyurea (HU) at 100 mM, and 4-nitroquinoline 1-oxide (4NQO) at 0.2 M.
into a synchronous cell cycle at non-permissive temperature for the smc5-3KE mutant (i.e. 18°C). Samples were taken at 30-min intervals, and cell cycle progression was monitored using flow cytometry and bud morphology. Yeast carrying the tub2-401 allele or an epitope-tagged version of wild-type SMC5 (i.e. the STH epitope that is also present in the smc5-3KE allele (17)) were included as controls. Wild-type cells and SMC5-STH controls progressed normally in the cell cycle, accumulating small buds concomitantly with the initiation of DNA synthesis (at 90 min after release; Fig. 8A) and entering M phase 30 -60 min thereafter (as evidenced by the increased population of large budded cells in the culture; Fig. 8A). The subsequent appearance of unbudded cells and return to a 1n genomic DNA content marked an exit from mitosis and return to G 1 210 min after the release (Fig. 8, A and B). In parallel experiments, the tub2-401 and smc5-3KE mutant populations showed similar kinetics of DNA replication and mitotic entry compared with wild-type cells (Fig. 8B). However, at later time points, the two mutant populations failed to exit mitosis and return to G 1 with normal kinetics. Indeed, tub2-401 and smc5-3KE cells accumulated with a 2n/large budded phenotype and remained in this state until the end of the experiment (Fig. 8, A and B), suggesting a persistent mitotic arrest in the absence of the TUB2 and SMC5 functions. A small fraction of the smc5-3KE mutant population rebudded toward the end of the experiment (i.e. at 240 min), whereas tub2-401 mutant cells remained mostly arrested with no bud under identical conditions (Fig. 8A). The population of rebudded smc5-3KE mutants may represent cells that have adapted to the spindle checkpoint arrest, as observed previously (45).
The mitotic arrest phenotype of smc5-3KE cells is similar to that of tubulin mutants (46), raising the possibility that Smc5 might affect microtubule functions in mitosis. To test this possibility,wemonitoredmicrotubulemorphologybyimmunofluorescence in SMC5 and smc5-3KE yeast cultures progressing synchronously in mitosis (Fig. 9A). Although smc5-3KE cells were clearly capable of building a microtubule cytoskeleton, our analysis revealed a significant proportion of mitotic cells with aberrant or defective microtubule morphology at 18°C (Fig. 9B). Moreover, there was a noticeable fraction of smc5-3KE mutants carrying anaphase spindles with unsegregated nuclear masses, indicating a defect in microtubule-driven separation of chromosomes (Fig. 9, B and C). Furthermore, relative to wild-type controls, a significant fraction of smc5-3KE cells contained nuclear microtubules that differed from their normal morphology, with spindles containing more than one continuous nuclear structure (referred to as "branched"; Fig. 9B) by immunofluorescence. Likewise, another fraction of smc5-3KE cells contained spindles that were either mispositioned or misoriented, being entirely localized in the daughter cells or having initiated anaphase in the mother cell instead of at the bud neck (Fig. 9, B and C). Collectively, these results indicate that inactivation of the microtubule binding activity of Smc5 gives rise to defects in mitotic spindle morphology and positioning in vivo.

DISCUSSION
The ability of cells to maintain genomic integrity from generation to generation depends on faithful segregation of chromosomes. This is achieved, in large part, by ensuring that chromosomes are properly connected to the spindle microtubules during mitosis and meiosis. Proteins linking chromatin to microtubules play a crucial role in this respect (9,47). Here, we uncover the existence of an unexpected new activity in a known regulator of chromosome stability, the ability to bind both microtubules and DNA. This, together with previous mass spectrometry characterization of the microtubule interactome (13,15), suggests that Smc5 and other SMC family members represent a novel class of evolutionarily conserved proteins that can directly connect spindle microtubules to chromosomes during cell division. Furthermore, we provide evidence that this novel activity can contribute to the functional stability of microtubules and help sustain robust cell cycle progression in vivo.
The unique cold-sensitive phenotype of the smc5-3KE mutant relative to other alleles affecting Smc5-Smc6 complex components raises an interesting question: why a few point mutations in a single SMC protein have such a profound effect on microtubule stability? This question is particularly relevant in light of the fact that proteomic studies have shown that there are hundreds of proteins in the microtubule-associated proteome (13)(14)(15)(16). One explanation for this is that not all proteins in this group are likely to be direct regulators of microtubule dynamics and/or stability. In particular, we think that the phenotype of smc5-3KE mutant reflects the fact that budding yeast is a highly sensitized system to observe microtubule-related defects. Indeed, this organism has an unusually low concentra-   7 Because low tubulin concentration will favor microtubule depolymerization rather than its polymerization, yeast may be particularly reliant on microtubule stabilizers to build a functional spindle structure. Separate from a role in stabilization, SMC proteins, such as those of the Smc5-Smc6 or cohesin complexes, could provide anchorage points on the chromosomes for spindle microtubules to attach during mitosis. Given the high affinity of SMC proteins for microtubules, we postulate that SMC proteins are key stabilizing factors for microtubule polymers. Therefore, impairment of one major component in this system could have a noticeable consequence on mitotic spindle integrity, especially when cells are exposed to microtubule stressors, such as low temperature (41,42).  Mechanistically, how might Smc5 contribute to microtubule stability? We think this could be achieved simply through high affinity association with microtubules. The buffer condition that we used for measuring k d has much higher ionic strength than what was traditionally used for a microtubule-binding experiment, suggesting a very strong interaction between SMC proteins and microtubules. In fact, under this condition, most kinesins would barely interact with microtubules (29,52,53). Additionally, Smc5 might have multiple microtubule-binding sites and thus allow bundling of adjacent microtubules, as we have observed in our microscopy experiments (Fig. 4D). However, it remains unclear how much this would contribute to spindle integrity in yeast and whether it could partly account for the occurrence of the "branched" phenotypes observed in cells expressing smc5-3KE (Fig. 9, B and C). This is further complicated by the presence of other microtubule-regulating fac-tors in vivo. Given the small size of the yeast spindle, we could not resolve microtubule bundles (or the lack of them) by fluorescence microscopy. Separate from this bundling activity, high affinity binding by Smc5 could physically hinder tubulin disassembly at microtubule ends, which would effectively slow down catastrophe and promote rescue.
Functionally, besides preserving spindle integrity, we have considered other impacts of the newly described SMC-microtubule interaction. It is conceivable that microtubule binding could affect other Smc5 functions. For instance, this binding event could act to allosterically regulate the affinity of Smc5 with other binding partners, such as chromatin or other members of the Smc5-Smc6 complex (3,54). Alternatively, microtubule binding could sequester the pool of free Smc5 or Smc5-Smc6 complexes in cells and in such a way modulate the dynamics and availability of Smc5 for specific functions in cells, such as DNA repair (55). This is consistent with our observation that smc5-3KE is sensitive to DNA-damaging agents, although the precise regulatory mechanism in this process has yet to be determined.
Our work has uncovered a novel and unexpected function of SMC proteins in the regulation of microtubule stability. This discovery raises many interesting questions. For example, are other chromatin binding proteins also contributing to this function? If so, do these proteins collectively contribute to the formation and maintenance of microtubule-based structures, such as the mitotic spindle? In this regard, it is interesting to note that SMC proteins, including the Smc5-Smc6 complex, have been shown to be enriched at centromeres in yeast, suggesting a possible link between this complex and the regulation of kinetochore function (55)(56)(57)(58)(59). It is tempting to speculate that SMC proteins can take part in promoting kinetochore-microtubule interactions by capturing and stabilizing microtubules in close proximity of the centromeres. Regardless, our data suggest that SMC proteins could provide a linkage between chromatin and microtubules, and this could in turn directly contribute to faithful segregation of chromosomes during mitosis. In light of our finding, it will also be interesting to determine in the future whether microtubules can have a direct effect on the known functions of the SMC complexes, including chromosome cohesion, condensation, and DNA repair.