SART1 localizes to spindle poles forming a SART1 cap and promotes centrosome and spindle assembly

SART1 is overexpressed in various cancers. However, its physiological function and cancer relevance remains elusive. Here we identify SART1 as a mitotic-specic and Ran-regulated microtubule-associated protein. SART1 downregulation in human cells as well as its depletion from frog egg extracts disrupts spindle assembly. While SART1 is nuclear in interphase, it localizes during mitosis to spindle poles in a microtubule-dependent manner. SART1 accumulates close to centrosomes forming a half circle which we designate as SART1 cap. Immunoprecipitation of SART1 identies the centrosome scaffold protein Cep192 as an interaction partner. Accordingly, Cep192 downregulation abolishes SART1 localization to spindle poles, and SART1 downregulation displaces centrosomal proteins like Ninein from centrosomes, but does not affect γ-tubulin localization. Furthermore, SART1 downregulation selectively kills cancer cells and prevents normal cells from oncogenic transformation. Our data unravel a novel function of SART1 for centrosome organization and spotlight SRAT1 as a potential target for anticancer therapies. 2 µM importin a, 2 µM importin b, and 5 µM RanGTP for 15 min. The samples were centrifuged at 100,000 g for 10 min at 20°C, and pellets were incubated with CSF-XB supplemented with 500 mM NaCl for 5 min, and centrifuged again. The supernatant (eluate) was analyzed by Western blotting.


Introduction
The squamous cell carcinoma antigen recognized by T-cells 1 (SART1) gene was rst identi ed and cloned as a carcinoma antigen recognized by cytotoxic T-cells 1 . Indeed, the SART1 protein is expressed in most proliferating cells and is especially overexpressed in cancer cells including epithelial cancers 2 .
Therefore, SART1 has been considered as a potential diagnostic marker and therapeutic target in various cancers 3,4 . However, it is still unknown whether SART1 overexpression is a cause or a consequence of malignant cell transformation and the physiological function of SART1 remains elusive.
In cell free assays using HeLa cell nuclear extracts, SART1 is required for U4/U6.U5 tri-snRNP recruitment to the pre-spliceosome in the spliceosome assembly pathway and plays a role in pre-mRNA splicing 5 .
Therefore, SART1 is also called as U4/U6.U5 tri-snRNP-associated protein 1. In addition, SART1 was identi ed in genome wide RNAi screenings as one of the genes required for mitotic progression 6,7 .
Speci c RNAi screens found SART1 as critical factor for centriole biogenesis 8 and sister chromatid cohesion 9 . Furthermore, SART1 knockout in mice is embryonic lethal 10 . All this suggest that SART1 has crucial functions in mitosis and cell division with vital importance for organismic homeostasis and/or development. However, we still lack detailed knowledge about the molecular action of SART1 in mitosis.
In eukaryotic cells, proteins containing nuclear localization signal (NLS) reside inside the nucleus in interphase with critical nuclear functions, including DNA replication, transcription, and splicing. Upon nuclear envelope breakdown at the beginning of mitosis, however, nuclear proteins become accessible to cytoplasmic proteins, including microtubules (MTs), and can play time-speci c, moonlighting roles for spindle assembly and function 11 . Also, many of the nuclear proteins speci cally function in the vicinity of mitotic chromatin, because of the GTP-bound form of the Ran GTPase (RanGTP) produced locally around chromatin 12 . RanGTP binds to the heterodimeric nuclear transport receptor importin α/β and dissociates NLS-containing nuclear proteins from the importins. The liberated nuclear proteins are active and induce spindle assembly around chromatin. To identify such nuclear MT regulators, we have a nitypuri ed NLS-containing proteins and MT-associated proteins (MAPs) sequentially, and obtained over 200 proteins 13,14 . The NLS-MAPs has been proven to be an excellent resource to uncover new mitotic regulators and reveal their functions 15,16 .
One of the NLS-MAPs is SART1, a so far unrecognized MAP that uniquely localizes to spindle poles. We show here that SART1 is required for mitotic progression and spindle assembly. The SART1 function at the spindle pole involves its interaction with Cep192 and recruitment of some centrosomal proteins.

Results
SART1 is a bona-de microtubule-associated protein.
To test whether SART1 can indeed interact with MTs, we added taxol-stabilized MTs to HeLa nuclear extracts. Endogenous SART1 was e ciently co-sedimented, which indicates its MT binding ( Fig 1A).
Addition of recombinant importin a/b complex inhibited SART1-MT interaction ( Fig 1A). Inhibition was reversed by the co-addition of RanGTP, which binds to importin b and removes the importin complex from NLS-sites. As previously reported, the MT polymerase chTOG, the orthologue of Xenopus XMAP215, showed no inhibition by importins 16 . Similarly, SART1 was co-sedimented from Xenopus egg extracts with taxol-stabilized MTs (Fig S1A).
To examine whether SART1 directly interacts with MTs, we expressed recombinant Xenopus SART1 in bacteria, using an Acidic-Target Tag 17 for better solubility. In MT sedimentation assay, recombinant SART1 directly bound to MTs (Fig. S1B). This MT binding was inhibited by importin a/b and restored by further addition of RanGTP (Fig. S1C). These results indicated that SART1 is a bona-de, previously unrecognized, MAP, which is regulated by importins and RanGTP. SART1 is required for mitotic progression and spindle assembly.
To assess SART1 impact on mitosis, we knocked down SART1 in HeLa cells by RNA interference (RNAi). SART1 expression was e ciently down-regulated with three independent small-interfering RNA (siRNA) oligos ( Fig S2A). Live cell imaging of HeLa cells stably expressing histone H2B-mCherry was carried out with 3 min time frame (Fig 1B), and the imaging data was automatically analyzed using the CellCognition software 16,18 . Upon SART1 depletion, we found a signi cant delay of prometaphase progression and often saw that chromosomes do not properly align to the metaphase plate (Fig 1B and C). For further detailed characterization of SART1, we used siSART1#3 oligo, if not stated otherwise, because this oligo induces milder spindle defects and less cell death compared to the other two ( Fig. S2A-C and S3C).
Immunostaining of the xed siSART1#3-treated cells showed consistently chromosome misalignments and also spindle MT defects (Fig 1D and E). MTs appeared not tightly arranged inside the spindle, and spindle poles were not well focused. Similar spindle defects were observed with all of the three siRNAs (Fig. S2B). In agreement, we found by FACS analysis that SART1 depletion signi cantly induces apoptotic cell death (Figs. S2C and S3). Detail analysis of the live cell imaging data revealed that a considerable number of cells die as consequence of early mitotic defects (Fig. S3).
To further validate that the spindle defects are caused by the lack of SART1, we performed rescue experiments. We used a Sendai virus (SeV) vector system to transiently express a SART1 version resistant to downregulation by siSART1#3 oligo 19 (Fig S2D). Indeed, the exogenous SART1 was well expressed in HeLa cells and fully rescued the spindle defects ( Fig. 1 D-F).
Chromosome attachment defects and spindle checkpoint activation in SART1-depleted cells.
To understand the nature of the spindle defects, we immunostained SART1-depleted cells for the kinetochore marker CREST. In control cells, kinetochore pairs were connected to MTs emanating from both spindle poles with end-on attachment (Fig. 1G). In contrast, in SART1 depleted cells we could not detect clear MT-kinetochore interactions (Fig. 1G). Functional end-on attachments are known to generate tension to separate kinetochores of sister chromatids, increasing the interkinetochore distance. We costained the cells for CREST and CENP-A, a histone H3 variant incorporated in centromeric DNA, and measured the distance of two CENP-A dots. Whereas control cells showed an inter-kinetochore distance of 1.1 μm, in SART1 depleted cells the distance was reduced to 0.8 μm (Fig. 1H). Staining with an outer kinetochore marker Ndc80/Hec1 also showed shorter interkinetochore distance upon SART1 downregulation (Fig. S2E). Consistently, the mitotic checkpoint protein BubRI remains localized at the metaphase spindle (Fig. S2G), indicating activation of the spindle checkpoint in the absence of SART1.These results indicate that SART1 is required for establishing and/or maintaining MT-kinetochore attachment and suggest that sister chromatides remain connected upon SART1 downregulation. Indeed, on chromosome spreads, we observed proper sister chromatid cohesion, at least in our depletion condition (Fig. S2F). SART1 localizes near centrosomes in the presence of dynamic microtubules.
To understand the place where SART1 functions, we examined the localization of SART1 in HeLa cells. A human SART1 antibody stained the nucleus and partially the cytoplasm in interphase cells ( Fig. 2A) as reported 20 , but surprisingly labeled the spindle poles in mitosis (Fig. 2B). This staining is speci c because it was lost upon SART1 downregulation ( Fig. 2A and B). After nuclear envelope breakdown, SART1 partially localized to the centers of two MT asters at prophase and accumulated further to the arising spindle poles at prometaphase (Fig. S4A). The spindle pole localization was most prominent in metaphase, remained through early anaphase and disappeared in telophase (Fig. S4A).
The major component of spindle pole is a centrosome that functions as the primary MT-organizing center. It consists of two centrioles surrounded by the pericentriolar material (PCM). When costained with a PCM marker γ-tubulin, SART1 localized around γ-tubulin in early mitosis, but interestingly formed a half circle around the γ-tubulin facing away from the spindle axis (Fig. 2C). The SART1 signal was similarly detected as a half circle around another PCM marker, pericentrin (Fig. 2D). This SART1-pericentrin con guration was also observed using a different SART1 antibody (SART1 #2) (Fig. S4B).
Given the identi cation of SART1 as a MAP, we examined how MTs affect the spindle pole localization of SART1. Before xation, HeLa cells were treated for 10 min with taxol to stabilize MTs or with nocodazole to depolymerize MTs. In both conditions, SART1 disappeared from the spindle poles (Fig. 2E), indicating a requirement of dynamic MTs for the SART1 localization. This contrasts to the pole localization of γtubulin which is not displaced by taxol or nocodazole treatment (Fig. S4C). These results indicate that SART1 localizes to the distal surface of centrosomes along the spindle axis and that this localization is dependent on dynamic MTs. SART1 interacts with Cep192, which promotes the pole accumulation of SART1.
To understand the mechanism of how SART1 localizes near centrosomes, we immunoprecipitated endogenous SART1 from Xenopus egg extracts using a speci c antibody we generated ( Fig. 3A and S5A). SART1-interacting proteins were eluted from the antibody beads with a high pH buffer (Fig. S5B), while SART1 remained mostly on the beads (Fig. S5B) presumably due to strong antigen-antibody interaction. Analysis of the eluate by shotgun mass spectrometry identi ed known interacting partners of SART1 involved in mRNA splicing, for example SF3B3 (Table S1). As expected, SART1 itself was found with low score (quantitative value 1).
Intriguingly, mass spectrometry identi ed a number of centrosomal proteins, containing Centrosomal protein of 192 kDa (Cep192) with the third highest score (Table S1, quantitative value 92). Cep192 is an integral component of centrosomes regulating recruitment of many centrosomal proteins and is required for centriole duplication and centrosome maturation 21,22 . To con rm the SRAT1-Cep192 interaction, we reciprocally immunoprecipitated Cep192 from HeLa cell lysate using a commercial antibody. Although in our hand we could not detect Cep192 band by Western blot, we found SART1 in the Cep192 precipitates ( Fig. 3B). Co-staining of the two proteins showed that SART1 localizes in proximity to Cep192, with SART1 covering the distal surface of the Cep192 spots in metaphase (Fig.3C).
When Cep192 was depleted in HeLa cells with an established siRNA 23 , SART1 are was undetectable at spindle poles (Fig. 3D). Yet, SART1 protein levels were unchanged upon Cep192 depletion (Fig. 3E), indicating that Cep192 is required for accumulation of SART1 on spindle poles but not for its stability. In contrast, SART1 depletion resulted in only partial (~30%) reduction of Cep192 staining at mitotic centrosomes ( Fig. 3F and G). While we do not know protein level of Cep192 in the absence of SART1, we con rmed that mRNA level of Cep192 unchanged in the absence of SART1 (Fig. S5C). We conclude that Cep192 is a interaction partner of SART1 critically required for SART1 accumulation near mitotic centrosomes. In contrast, SART1 partially increases Cep192 levels at centrosomes. SART1 is required for recruitment of centrosomal proteins, throughout the cell cycle.
Since SART1 accumulates near mitotic centrosomes and interacts with centrosomal proteins, we examined the localization of other key centrosomal components upon SART1 depletion. The staining intensity of the MT anchoring protein Ninein 24 at centrosomes decreased in mitosis but surprisingly also in interphase (Fig. 4A), upon SART downregulation using three different SART1 siRNAs. The centrosome staining of the scaffolding protein pericentrin was also decreased throughout the cell cycle (Fig. S6A), but in this case the reduction in interphase was even more prominent (Fig. S6A). The staining of the centriolar satellite protein PCM1, also present in the SART1 immunoprecipitates (quantitative value 11), was dispersed both in mitosis and interphase (Fig. 4B) upon SART1 downregulation. We quanti ed this phenotype by measuring centrosomal versus cytoplasmic staining intensity of PCM1 (Fig. 4B), which showed PCM1 mislocalization both in interphase and mitosis. Similarly, EB1, a MT plus-end binding protein known to localize to centrosomes 25 , also showed reduced staining at mitotic and interphase centrosomes in SART1-depleted cells (Fig. 4C) indicating a less e cient EB1 recruitment to centrosomes.
Total EB1 protein level were unchanged upon SART1 downregulation (Fig.S2), indicating inhibition of EB1 recruitment to centrosomes. Also, mRNA levels of Ninein, Pericentrin, and PCM1 were not reduced in the absence of SART1 (Fig. S6B).
Reduction of centrosomal protein localization in interphase could indicate altered MT functions. MTs regulate cell polarization and migration in interphase 26 . We thus tested these functions using an in vitro wound healing assay. siSART1-treated U2OS cells showed clear reduction of cell migration ( Fig. 4D) pointing to a role of SART1 also for interphase MT function.
Whereas centrosome accumulation of Ninein, pericentrin, PCM1 and EB1 was impeded by SART1 downregulation, recruitment of γ-tubulin, a primary MT nucleating factor, was not affected (Fig. S6C). This is different to the phenotype observed upon Cep192 downregulation which prevents γ-tubulin centrosome localization 21 . Also, distinct from Cep192 22 , depletion of SART1 did not affect the number of centrioles, labeled with Centrin, at the spindle poles ( Fig. S6D. Thus, SART1 is required to recruit selected proteins to the centrosome throughout the cell cycle, but this function is likely independent of Cep192. SART1 promotes spindle bipolarity in Xenopus egg extract via its N-terminal NLS-containing region. To further understand the molecular function of SART1, we used Xenopus egg extracts. In this cell-free system, spindle assembly can be faithfully recapitulated when sperm heads are added to the extracts and cell cycle activation is trigged into interphase and then to mitosis. SART1 antibody, used for the immunoprecipitation (Fig. 3A), e ciently depleted endogenous SART1 from egg extracts (Fig. 5A). Whereas control-treated (mock) extracts assembled bipolar spindles with chromosomes aligned at the metaphase plate, depletion of SART1 caused defects in spindle assembly (Fig.5B). Although MT polymerization around sperm chromatin was equivalent to the control situation, bipolar spindle orientation was not established and chromosomes did not align to the center (Fig.5B, C, and E). These spindle phenotypes were rescued by addition of Xenopus SART1 mRNA to express recombinant SART1 (Fig. 5A, B, and E).
To understand which parts of SART1 are crucial for its spindle function we constructed N-and C-terminal deletion mutants of the protein (Fig. 5D) and expressed them in SART1 depleted egg extracts (Fig. 5E). Besides the full-length SART1, the fragment lacking the last hundred aa at C-terminus (aa1-665) restored bipolar spindle assembly (Fig. 5E). In contrast, mutants lacking N-terminus did not, which suggests that a potential NLS (aa 26-46 of Xenopus SART1, predicted by cNLS Mapper 27 ) in the N-terminus might be important for MT binding and spindle bipolarization.

SART1 depletion induces spindle defects in cancer cells, but not in normally growing cells
Because SART1 has been implicated in tumorigenesis 3,4 we overexpressed SART1 or an oncogene c-Myc 28 in non-transformed diploid retinal pigment epithelial 1 (RPE1) cells using SeV vector carrying SART1 (Fig. S2E) or c-Myc. After 1 month with occasional medium changes, control and SART1expressed cells grew in a monolayer and stopped further growth due to contact inhibition 29 , while c-Mycexpressed cells grew in nitely, became smaller, and formed multilayered colonies (Fig. S7A) 28 . The result suggests that SART1 is not a cause of tumorigenesis.
Next, we downregulated SART1 expression in normal and cancer cell lines. siSART1 treatment similarly downregulated SART1 in normal BJ skin broblasts and RPE cells, as well as in osteosarcoma-derived U2OS cells (Fig. 6A). Immunostaining indicated that SART1 depletion induces spindle and chromosome alignment defects only in U2OS cells (Fig. 6B, C, and S7B). Consistently, SART1-depleted RPE1 cells did not show increased apoptotic cell death (Fig. S7B), whereas SSX2IP depletion, which induces spindle defects in RPE1 cells 30 , induced cell death (Fig. S7C).
RPE1 and BJ have an elongated cell shape, distinct from round morphology of cancer-derived HeLa and U2OS cells, and these cultures contain a lower fraction of mitotic cells than HeLa and U2OS cell cultures 31 . To examine whether these differences affect cell response to SART1 depletion, we used iPS (induced pluripotent stem) cells as non-cancerous cells with a high mitotic index 32,33 . E cient SART1 downregulation in iPS cells neither affected spindle assembly nor caused cell death (Fig. 6D).
These experiments indicate that upon SART1 depletion, the cancer cell lines show a higher rate of cell death as compared to the normal cell lines. To examine if cell transformation changes their response to SART1 depletion, we downregulated SART1 expression in RPE1 cells and transformed them by c-Myc expression using the SeV vector carrying c-Myc. Cells were cultured for total 2 weeks with siRNA/medium change every 3 day. As expected, control RPE1 cells stopped growing while c-Myc-transformed cells grew in nitely (Fig. 6E). Interestingly, SART1 downregulation strikingly inhibited c-Myc-induced cell transformation without affecting control cells ( Fig. 6E and S7D). This result, together with no cell death induced in SART1-depleted RPE1 cells (Fig. S7), indicates that SART1 downregulation preferentially induces cell death in transformed cells. Discussion SART1 has been previously identi ed in RNAi screenings as a protein required for mitotic progression 6,7 . It was also found in a RNAi screening searching for proteins required for centrosome biogenesis 8 . However, mitotic and centrosomal function of SART1 has not been characterized in detail and it remained unclear whether SART1 is required for those events directly or indirectly via pre-mRNA splicing of relevant factors.
Here we demonstrate SART1 as a bona de RanGTP-regulated MAP. SART1 localizes next to centrosomes and is required for centrosome maturation and spindle assembly. Some splicing factors (Cdc5l and the Cdc5l-containing Prp19 complex) have been shown to be important for spindle assembly 34,35 . Although these proteins do not localize to speci c mitotic structures, their functions on spindle assembly are independent of mRNA splicing 35 . Thus, our data together with previous evidence support the idea that multiple splicing factors have moonlighting functions in mitosis 11 . SART1 has been described as one of the pre-mRNA splicing factors required for sister chromatid cohesion 9 . We show, however, by chromosome spreads that SART1-depleted cells exhibit proper sister chromatid cohesion. SART1 depletion also causes chromosome misalignment with reduced interkinetochore distance. This is distinct from the phenotype of cohesin component depletion which results in longer interkinetochore distances 36 . Both results indicate that sister chromatid cohesion is normal, at least under our depletion condition. Therefore, potential SART1 cohesion functions are unlikely to cause the spindle and centrosome assembly defects we observed.
SART1 resides inside of nucleus in interphase and is during his stage of the cell cycle undetectable at centrosomes. Upon nuclear envelope breakdown, SART1 starts accumulating around the two centrosomes and further increases in prometaphase. As bipolar spindles establish during metaphase, SART1 forms a half circle around centrosomes averted from the spindle axis. The SART1 localization near centrosomes is dependent on the presence of dynamic MTs, suggesting that it is not an intrinsic component of centrosomes. Intriguingly, the SART1 and PCM proteins, such as γ-tubulin and pericentrin, localize in close proximity but do not overlap. We are not aware of another protein with a similar localization pattern as SART1 at the centrosome periphery. Therefore, we propose this novel centrosomal structure as the SART1 cap, which is potentially important for centrosomal maturation and polar spindle formation. How this cap-like structure forms is an interesting question for the future. We envision that forces moving SART1 toward the cell cortex are important for establishment of the structure.
We found by immunoprecipitation that SART1 interacts with Cep192, the scaffold of centrosome assembly 21,22 . Cep192 depletion in HeLa cells abolishes SART1 accumulation at the spindle poles, whereas SART1 downregulation only partially decreases Cep192 levels at centrosomes. Depletion of either SART1 or Cep192 reduces centrosome staining of several PCM and satellite components. However, whereas Cep192 depletion abolishes γ-tubulin staining at mitotic centrosomes 21,22 , SART1 depletion does not impact centrosome localization of this MT nucleation factor (Fig. S6B). Cep192 is known to have dual roles in centrosome duplication and maturation 22 . However, it remains unclear to which extent Cep192 contributes to each function because null mutation or RNAi depletion of Cep192 causes loss of the entire organelle 37 . In contrast to Cep192, depletion of SART1 does not affect the number of centrin dots, indicating that SART1 functions in centrosome maturation but not duplication. These results indicate that although SART1 requires Cep192 for its localization near centrosomes, the downstream functions of SART1 at centrosomes are distinct and probably complementary to that of Cep192.
By immuno uorescence, SART1 is not detectable at centrosomes in interphase but accumulates around centrosomes upon mitotic entry. It is, therefore, surprising that SART1 depletion impairs the localization of centrosomal proteins not only in mitosis but also in interphase. For Ninein, PCM1, EB1 the degree of reduction or scattering is comparable between interphase and mitosis, and even more pronounced reduction of centrosomal staining intensity is observed in interphase for pericentrin. However, SART1 depletion does not reduce mRNA levels of Ninein, PCM1, and Pericentrin, and affect the protein level of EB1, indicating that it does not impair mRNA splicing or protein stability. Thus, it is currently unknown how SART1 regulates their localization but we envision two possible scenarios. First, given the partial localization of SART1 in the interphase cytoplasm 20 (Fig. 2A), small amounts of SART1 beyond the immuno uorescence detection limit may be present at centrosomes in interphase and recruit the centrosomal proteins. Second, SART1 at mitotic centrosomes may act as licensing factor for recruitment of Ninein, PCM1, EB1 and pericentrin in the subsequent interphase.
Similar to RNAi-based knockdown of SART1 in human cells, immunodepletion of SART1 in Xenopus egg extracts causes spindle assembly and chromosome alignment defects, without affecting MT amounts assembled around chromatin. These defects are rescued by expression of exogenous SART1, showing the speci city of the observed phenotypes. Compared to the experiments in human cells, SART1-depleted extracts show more severe defects in spindle pole assembly. This might re ect the larger size (~40 μm in length) of spindles assembled in extracts which results in stronger phenotypes at the places where the depleted protein functions 13 . Alternatively, it might be due to very e cient and rapid depletion setting in egg extracts versus a time demanding downregulation with potential adaptations in cells. Using egg extracts, we further identify the N-terminus of Xenopus SART1 as essential for rescueing the spindle defects. The N-terminus contains a predicted NLS (aa 26-46 of Xenopus SART1 27 ). Considering that several nuclear proteins bind to mitotic MTs via their NLS and neighboring redidues 38 , we propose that the N-terminus of SART1 is required for spindle pole formation via its MT binding region. SART1 is overexpressed in various cancer cells 2 but it has been unclear whether the overexpression is a cause or a consequence of tumorigenesis. Our experiments overexpressing SART1 in RPE1 cells did not cause cell transformation, whereas c-Myc expression using a comparable SeV vector transforms these cells. This supports the idea that SART1 overexpression occurs as a consequence of tumorigenesis by re ecting active cell proliferation 2 . On the other hand, downregulation of SART1 e ciently induces programmed cell death in cancer cells but not in normal cells. Upon SART1 downregulation, a considerable number of HeLa cells die after spindle assembly/chromosome alignment defects. Furthermore, SART1 depletion prevents hyperproliferation of RPE1 cells after c-Myc-induced transformation. We imagine that the depletion induces cell death only when RPE1 cells are transformed and this selective cell death prevents tumorigenic-like behavior of cells. It will be interesting and important to understand how cell death is speci cally induced in cancer cells. Given that SART1-depleted egg extract shows severe spindle assembly defects and SART1 knockout mice are embryonic lethal 10 , differentiated healthy cells may have redundant mechanisms that compensate the loss of SART1.
Alternatively, since cancer cells preferentially die with antimitotic drugs taxol and nocodazole 31,39 , they may be sensitive to the downregulation of SART1 required for spindle assembly. In either case, SART1 depletion or inactivation have an impact on cancer proliferation and should be considered as therapeutic strategy.
In summary, our results unravel SART1 as a novel MAP that localizes near centrosomes forming a caplike structure. Functionally, SART1 promotes recruitment of speci c centrosomal components for centrosomal maturation and spindle pole assembly. In addition, depletion of SART1 causes cell death selectively in cancer cell lines establishing SART1 as an attractive target for cancer therapy.

Recombinant proteins and antibodies
A cDNA covering the full-length Xenopus laevis SART1 (NM_001086027.1) was in vitro synthesized (GenScript) and subcloned into pET-28a (Novagen) with NdeI and XhoI sites. For antibody production, the recombinant Xenopus SART1protein was expressed in BL21 (DE3) E. coli and solubilized from inclusion bodies with 6 M urea. The protein was puri ed with Ni-NTA Agarose (Qiagen), dialyzed to PBS containing 6 M urea, and used for immunization in rabbits (Hokkaido System Science). From the antisera, Xenopus SART1 antibody was puri ed using the antigen column and used for Western blot at 1 mg/ml. To increase recombinant SART1 solubility, His-Acidic-Target Tag (HisATT), modi ed from Flag-Acidic-Target Tag 17 , was synthesized (Genscript) and subcloned into the above mentioned pET-28a based plasmid with NcoI and NdeI sites. The HisATT-SART1 was expressed in E. coli and puri ed with Ni-NTA, and dialyzed to CSF-XB buffer (10 mM K-HEPES, 100 mM KCl, 3 mM MgCl 2 , 0.1 mM CaCl 2 , 50 mM sucrose, and 5 mM EGTA, pH 7.7) containing 10% glycerol and 1 mM DTT, and used for MT sedimentation assay. For in vitro translation in egg extracts, Xenopus SART1 and its depletion mutants were made by PCR and subcloned into a pCS2+ vector harboring FLAG tag sequence at the N-terminus. Importin a, importin b, and RanQ69L-GTP were expressed in E. coli and puri ed with TALON beads 15 .
MT binding assays 0.1 µM recombinant SART1 was incubated with 2 µM taxol-stabilized MTs for 15 min, and centrifuged at 20,000 g in a TLA120.2 rotor (Beckman) for 10 min at RT. The Supernatant and pellet was analyzed by Coomassie staining or Western blot. The assay was also done in the presence or absence of recombinant 3 mM importin a, 3 mM importin b, and 5 mM RanQ69L-GTP, a dominant positive mutant of Ran locked in the GTP-bound state.
HeLa nuclear extract (4C Biotech) was diluted to 1 mg/ml with CSF-XB buffer. HeLa nuclear extract (1 mg/ml) was incubated with 2 mM pure taxol-stabilized MTs in the presence or absence of recombinant importin a/b complex and RanGTP, and pelleted. Microtubule-associated proteins (MAPs) were eluted with high salt from the pellet, and the supernatant after a second centrifugation was analyzed by immunoblot.
Xenopus CSF egg extracts were diluted 1:3 with CSF-XB buffer to a concentration of about 30 mg/ml.
After centrifugation in a TLA-100.2 rotor at 100,000 x g for 10 min at 4°C, the supernatant was incubated at RT in the presence or absence of 2 µM taxol-stabilized microtubules, 2 µM importin a, 2 µM importin b, and 5 µM RanGTP for 15 min. The samples were centrifuged at 100,000 g for 10 min at 20°C, and pellets were incubated with CSF-XB supplemented with 500 mM NaCl for 5 min, and centrifuged again. The supernatant (eluate) was analyzed by Western blotting.
For rescue experiments, HeLa cells were infected with SeVs at MOI3. After 8 h incubation, medium was replaced to fresh one containing 20 nM siRNAs, and cells were cultured for additional 72 h.
When indicated, HeLa cells were treated with 10 μM taxol or 20 μM nocodazole for 10 min, and xed with Mildform.
For immuno uorescence staining, cells were incubated with blocking buffer (PBS + 2% BSA + 0.1% Triton-X100) at RT for 30 min, and then incubated with the primary antibodies in the blocking buffer at 4°C overnight. Cells were washed with PBS, incubated with the secondary antibodies and DAPI at RT for 30 min, washed again with PBS, and mounted with Mowiol 4-88 (Calbiochem).
Fluorescence images were acquired by using an Olympus Fluoview FV1000 confocal microscope equipped with a UPlanSApo 60x/1.35 Oil objective at 0.5 μm Z steps. Maximum intensity projections were obtained using ImageJ (NIH). Confocal slices are used to detect kinetochore-MT attachment and sister kinetochore pairs. Interkinetochore distance was measured using ImageJ. Maximum projected images are used to quantify uorescent intensity using ImageJ. Accumulation of PCM1 at spindles poles was calculated as ratio of PCM1 signal intensity at poles (two 3 μm circles) and larger area around chromatin (10 μm circle) (Fig. 4). Accumulation at interphase centrosomes was calculated as ratio of PCM1 signal intensity at centrosomes (3 μm circle) and larger area around centrosome (10 μm circle) (Fig. 4).
For cell transformation assay, RPE1 cells were infected with SeV harboring SART1 or c-Myc at MOI3 overnight, replaced with fresh medium, and cultured for 1 month with occasional medium change. After xation with Mildform 10N, cells are imaged by Leica DMIRB bright-eld microscope equipped with a N Plan 5x/0.12 objective. To examine the effect of SART knockdown on c-Myc-induced transformation, RPE1 cells were treated with siRNAs and 2 days later infected with SeV harboring c-Myc with MOI3. siRNAs and medium were replaced every 3 days. Cells were xed in 2 weeks after the rst RNAi treatment.
The wound healing assay was performed using culture-insert 2 well dishes (ibidi, 80206). After 48 h of siRNA treatment in U2OS cells, inserts were removed and cell migration to the vacant zone was imaged every 30 min using a Nikon ECLIPSE Ti2 phase-contrast microscope equipped with a Plan Apo 4x/0.20 objective, 37°C chamber, 5% CO2, and humidity control. Quantitative real-time PCR 3 day after siRNA transfection, total RNAs were isolated using PureLink RNA Mini Kit (Ambion) and were subjected to PureLink DNase (Invitrogen) treatment. Reverse transcription and quantitative real-time PCR were conducted using TaqMan RNA-to-Ct 1-step Kit and 7500 Fast Real-Time PCR System (both from Applied Biosystems) with the indicated primer and probe sets (Table S2). To amplify only spliced mRNAs, one of the primers or probe was designed on exon-intron junction. Values were normalized to those of GAPDH.
Xenopus egg extracts and cell-free assay Cytostatic factor-arrested M-phase Xenopus laevis egg extracts (CSF extracts) were prepared as described 41 . In short, Xenopus eggs were dejellinated by cysteine treatment, washed with CSF-XB buffer For spindle assembly in cycled extract 41 , CSF extract was supplemented with demembraned sperm, Cy3labeled tubulin, and 0.4 mM CaCl 2 , and incubated at 20°C for 90 min to allow cell cycle progression into interphase. Samples were cycled to mitosis by addition a fresh CSF extract and incubation at 20°C for 80 min. MT density around sperm was quanti ed using ImageJ. For rescue experiments, mRNAs for SART1 full-length and deletion mutants were prepared using mMESSAGE mMachine SP6 kit (Life technologies) and was added to the depleted extract at the beginning of the reactions.
Immunoprecipitation of SART1 and mass spectrometry Dynabeads Protein A were coupled with rabbit IgG or the Xenopus SART1 antibody following the manufacturer's instruction. The antibody beads were cross-linked with dimethyl pimelidate. Each bead sample (600 μl slurry) was incubated with Xenopus CSF egg extracts (1000 μl) at 4°C for 60 min, washed twice with CSF-XB and twice with CSF-XB containing 0.5 M KCl and 0.1% Triton X-100. The immunoprecipitates were resuspended in SDS sample buffer and resolved by SDS-PAGE for Coomassie stain or immunoblot.
For shotgun mass spectrometry, the immunoprecipitates (600 μl bead slurry) were eluted from the beads with 0.1 M triethylamine (pH 11.5, 60 μl), and neutralized by addition of nal 0.1 M Tris pH 6.8. The proteins were trypsin digested and analyzed with a Q-Exactive mass spectrometer (Thermo Fisher Scienti c) at the CoMIT in Osaka University. The peptide data were analyzed by Scaffold (Proteome Software Inc.) and quantitative value was obtained by normalizing total spectra with molecular weight.
For reciprocal immunoprecipitation, human Cep192 antibody (GeneTex) was coupled and cross-linked to Dynabeads Protein A as described above. HeLa cells were cultured to con uence, washed with PBS, and lysed with RIPA buffer (50 mM Tris pH 8.0, 150 mM NaCl, 50 mM sodium orthovanadate, 1% NP40, 0.1 mM PMSF) supplemented with complete protease inhibitor cocktail (Roche). After centrifugation, the total cell lysate (200 μl, ~5 mg/ml) was incubated with the antibody beads (50 μl slurry) at 4°C for 60 min. Beads were washed with the RIPA and resolved by SDS-PAGE for immunoblot.
Statistical analyses (describe whether one-or two-tailed) For live cell imaging, data were tested for normality by D'Agostino and Pearson omnibus test. For normal distributions, ANOVA test and subsequent Dunnett multiple comparisons test were applied. When normal distributions could not be assumed, statistical signi cance at alpha = 0.001 was determined using a Kruskal-Wallis test followed by Dunn ś multiple comparisons test.
For other xed cell and egg extract experiments, student´s t-test was performed using Microsoft Excel with two-tail distribution and two-sample equal variance.

Declarations Data Availability
All data of this study are stored at the corresponding author and available on reasonable request. (The source data for the gures are available as a Source Data le.) Figure 1 SART1 is a MAP required for early mitotic progression and spindle assembly. A. Human SART1 binds to microtubules (MTs) in a RanGTP-regulated manner. HeLa nuclear extract was incubated with taxolstabilized MTs in the presence or absence of recombinant importin a/b complex and RanGTP and pelleted. Microtubule-associated proteins (MAPs) were eluted with high salt from the pellet. After centrifugation, and the supernatant (the eluate) was analyzed by immunobloting. B. HeLa cells expressing mCherry-H2B were transfected with indicated siRNAs and imaged in intervals of three minutes. Orange lines indicate cells in prometaphase. C. SART1 downregulation extends prometaphase.

Figures
Duration of prometaphase was quanti ed for more than 100 mitotic cells per condition. Statistical signi cance at alpha = 0.0001 was determined using a Kruskal-Wallis test followed by Dunn´s multiple comparisons test. D. SART1 is required for spindle assembly. HeLa cells were infected with control Sendai virus (SeV) vector or SeV carrying an siRNA-resistant version of human SART1. The cells were subsequently treated with control or SART1 siRNAs for 3 days, xed, stained for α-tubulin and DNA, and imaged by a confocal microscope. Maximum-projected images are shown. E. Quantitation of abnormal spindles assayed in D. Error bars: SD. N = 4 experiments, n > 50 prometaphase and metaphase-like cells were counted per experiment. Note that prophase, anaphase, telophase-like cells were not considered. F. Western blot of lysates from HeLa cells assayed in D. G. MT-kinetochore interactions are compromised upon SART1 downregulation. HeLa cells were treated with the siRNAs, xed, and stained for the kinetochore marker CREST, α-tubulin, and DNA. Confocal slice images are shown. Note that end-on attachment of MTs to kinetochores are seen in control cells but could not be resolved upon SART1 downregulation. 3-fold magni ed images are shown below. H. interkinetochore distance is reduced upon SART1 downregulation. The RNAi-treated cells were stained for the centromere marker CENP-A, CREST, and DNA. Confocal slices were used to nd kinetochore pairs and measure interkinetochore distance. n > 94 kinetochore pairs. p values (t-test, two-tailed).Scale bars, 10 mm. SART1 localizes to a speci c subdomain of mitotic centrosomes. A and B. HeLa cells were treated with control or SART1 siRNAs for 3 days, xed, and stained for SART1 (red), α-tubulin (green), and DNA (blue). A. In interphase, SART1 localizes within nucleus. B. SART1 localizes to spindle poles. Note that SART1 signals disappear in SART1 RNAi cells, validating the SART1 localization found in control cells. C. HeLa cells were xed, and stained for SART1, γ-tubulin, and DNA. Note that SART1 localizes around γ-tubulin in early mitosis, but eventually forms a half circle around the γ-tubulin spot distant from the spindle axis. D.
HeLa cells were xed, and stained for SART1, pericentrin, and DNA. SART1 localizes outside of pericentrin in metaphase. E. The SART1 localization at spindle poles requires dynamic MTs. HeLa cells were treated with excess amount of taxol or nocodazole for 10 min before xation and stained for SART1 α-tubulin, and DNA. Scale bars, 10 mm.

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download.