Defining endogenous TACC3–chTOG–clathrin–GTSE1 interactions at the mitotic spindle using induced relocalization

A multiprotein complex containing TACC3, clathrin, and other proteins has been implicated in mitotic spindle stability. To disrupt this complex in an anti-cancer context, we need to understand the composition of the complex and the interactions between complex members and with microtubules. Induced relocalization of proteins in cells is a powerful way to analyze protein-protein interactions and additionally monitoring where and when these interactions occur. We used CRISPR/Cas9 gene-editing to add tandem FKBP-GFP tags to each complex member. The relocalization of endogenous tagged protein from the mitotic spindle to mitochondria and assessment of the effect on other proteins allowed us to establish that TACC3 and clathrin are core complex members and that chTOG and GTSE1 are ancillary to the complex, respectively binding to TACC3 and clathrin, but not each other. PIK3C2A, a membrane trafficking protein that binds clathrin, was previously proposed to also bind TACC3 and stabilize the TACC3–chTOG–clathrin–GTSE1 complex during mitosis. We show that PIK3C2A is not on the mitotic spindle and that knockout of this gene had no effect on the localization of the complex. We therefore conclude that PIK3C2A is not a member of the TACC3–chTOG–clathrin–GTSE1 complex. This work establishes that targeting the TACC3–clathrin interface or their microtubule-binding sites are the two strategies most likely to disrupt spindle stability mediated by this multiprotein complex.


Introduction
During mitosis, chromosomes are segregated with high precision to generate two genetically identical daughter cells. This segregation is driven by the mitotic spindle, a bipolar microtubule array with associated motor and non-motor proteins (Manning and Compton, 2008). One non-motor protein complex that binds spindle microtubules contains TACC3, chTOG (also known as CKAP5) and clathrin (Fu et al., 2010;Hubner et al., 2010;Lin et al., 2010;Booth et al., 2011). This complex is important for stabilizing the bundles of microtubules that attach to kinetochores (kinetochore-fibers, k-fibers) by physically crosslinking them (Booth et al., 2011;Hepler et al., 1970;Nixon et al., 2015Nixon et al., , 2017. Uncovering the molecular details of how proteins of this complex bind to one another and to microtubules is important to understand how i) mitotic spindles are stabilized and ii) we can target spindle stability in an anti-cancer context. During mitosis, phosphorylation of TACC3 on serine 558 by Aurora-A kinase controls the interaction between clathrin and TACC3 (Booth et al., 2011;Cheeseman et al., 2011Cheeseman et al., , 2013Hood et al., 2013;Burgess et al., 2018). This interaction brings together the N-terminal domain of clathrin heavy chain and the TACC domain of TACC3 to make the microtubule-binding surface (Hood et al., 2013). Despite having a microtubule-lattice binding domain, chTOG is not needed for the complex to bind microtubules and interacts with the TACC3-clathrin complex via its TOG6 domain, binding to a stutter in the TACC domain of TACC3 (Booth et al., 2011;Hood et al., 2013;Gutiérrez-Caballero et al., 2015).
Despite this detail, the exact composition of the complex on kinetochore microtubules is uncertain.
Recent biochemical evidence convincingly shows that GTSE1 binds clathrin's N-terminal domain and that this interaction localizes GTSE1 to spindle microtubules (Rondelet et al., 2020). Like chTOG, GTSE1 has the capacity to bind microtubules, but it appears to use TACC3-clathrin to bind the spindle (Monte et al., 2000;Scolz et al., 2012;Bendre et al., 2016). By contrast, PIK3C2A is a component of clathrin-coated vesicles where it acts as a lipid kinase (Gaidarov et al., 2001). It was recently proposed to act as a scaffolding protein in the TACC3-chTOG-clathrin complex by binding to both TACC3 and clathrin (Gulluni et al., 2017). PIK3C2A and GTSE1 bind to the same sites on clathrin's N-terminal domain (Gaidarov et al., 2001;Rondelet et al., 2020) and although clathrin has the capacity to bind multiple proteins (Smith et al., 2017;Willox and Royle, 2012), this raises the question whether the binding of PIK3C2A and GTSE1 to TACC3-clathrin at the spindle is mutually exclusive.
Dissecting this multiprotein complex is further complicated because each putative member is able to form subcomplexes that have different subcellular localizations (Gutiérrez-Caballero et al., 2015). TACC3-chTOG (without clathrin) localize to the plus-ends of microtubules (Nwagbara et al., 2014;Gutiérrez-Caballero et al., 2015). Similarly, GTSE1 binds plus-ends and can also stabilize astral microtubules of the mitotic spindle by inhibiting the microtubule depolymerase, MCAK/KIF2C (Scolz et al., 2012;Bendre et al., 2016;Tipton et al., 2017). PIK3C2A and clathrin are found in clathrin-coated vesicles away from the mitotic spindle (Gaidarov et al., 2001). Biochemical approaches do not have the capacity to discriminate these subcomplexes from the multi-protein complex on K-fibres. Therefore, subcellular investigation of protein interactions are required to answer this question. a protein to mitochondria in order to inactivate that protein (Robinson et al., 2010). In the original method, the target protein is depleted by RNAi and an FKBP-tagged version is expressed alongside MitoTrap (an FRB domain targeted to mitochondria), relocalization is achieved by the addition of rapamycin. This method has many advantages over slow inactivation methods such as RNAi knockdown or gene disruption (knock-out) approaches (Royle, 2013). We have previously used knocksideways in mitotic cells to investigate protein-protein interactions since any proteins that are in a complex with the target protein also become mislocalized to the mitochondria (Cheeseman et al., 2013;Hood et al., 2013). This approach has the added advantage that the subcellular location of proteins can also be tracked during the experiment, and that it can be done at specific times; allowing us to pinpoint where and when interactions occur. In this study, we applied a knocksideways approach to investigate how proteins of this complex bind to one another and to microtubules of the mitotic spindle.
Instead of overexpression and RNAi, we sought to tag each target protein with FKBP and GFP at its endogenous locus using CRISPR/Cas9-mediated gene editing. This strategy allows us to study these subcellular interactions at the endogenous level for the first time. The cell lines we have created are a multi-purpose "toolkit" for studying microtubule-crosslinking proteins by live-cell imaging, biochemistry, or electron microscopy (Clarke and Royle, 2018).

Results
Generation and validation of clathrin, TACC3, chTOG and GTSE1 knock-in HeLa cell lines. Our first goal was to tag four proteins with FKBP and GFP at their endogenous loci using CRISPR/Cas gene editing. Clathrin (targeting clathrin light chain a, LCa/CLTA), TACC3, chTOG/CKAP5, and the clathrin-interacting protein, GTSE1 were edited in HeLa cells so that they had a GFP-FKBP tag at their N-terminus or an FKBP-GFP tag at their C-terminus ( Figure 1A). The dual FKBP and GFP tag allows direct visualization of the protein as well as its spatial manipulation using knocksideways (Cheeseman et al., 2013;Robinson et al., 2010). Following editing, GFP-positive cells were isolated by FACS and were validated using a combination of PCR, sequencing, Western blotting ( Figure Figure S1B). These validation steps yielded a cell line for each protein that could be used for all future analyses. Homozygous knock-in was achieved for CLTA-FKBP-GFP, GFP-FKBP-TACC3, and GTSE1-FKBP-GFP. Despite multiple attempts to generate a homozygous knock-in for chTOG-FKBP-GFP, we only recovered heterozygous lines (more than twenty heterozygous clones in three separate attempts).
We assume that homozygous knock-in of chTOG-FKBP-GFP is lethal. The localization of tagged proteins in all cell lines was normal. In mitotic cells, clathrin is located on the spindle and the cytoplasm/coated pits, TACC3 is located exclusively on the spindle, chTOG is located on the spindle but more pronounced on the centrosomes and kinetochores, and GTSE1 is localized throughout the spindle and the cytoplasm ( Figure 1C, Supplementary Figure S1B), consistent with previous observations (Gergely et al., 2000(Gergely et al., , 2003Royle et al., 2005;Foraker et al., 2012;Bendre et al., 2016). Overexpression of TACC3 can result in the formation of aggregates (Gergely et al., 2000;Hood et al., 2013) which have recently been described as liquid-like phase-separated structures (So et al., 2019). We note that at endogenous levels in HeLa cells, GFP-FKBP-TACC3 does not form these structures ( Figure 1C, Supplementary Figure S1C). As a further validation step, we assessed mitotic timings in each knock-in cell line and found progression to be comparable to their respective parental HeLa cells. These observations indicate that the addition of an FKBP and GFP tag did not affect the mitotic function of clathrin, TACC3, chTOG, or GTSE1 and that clonal selection did not adversely affect mitosis in the four cell lines (Supplementary Figure  S1D). Finally, we assessed the functionality of the FKBP moiety by performing knocksideways experiments ( Figure 2A). Each cell line, expressing mCherry-MitoTrap was imaged live during the application of 200 nM rapamycin. At metaphase, CLTA-FKBP-GFP, GFP-FKBP-TACC3, chTOG-FKBP-GFP, and GTSE1-FKBP-GFP were all removed from the spindle and relocalized to the mitochondria by rapamycin addition ( Figure 2B). The time course of removal was variable but was complete by 10 min.
In summary, the generation and validation of these four knock-in cell lines represents a toolkit that we can use to study clathrin, TACC3, chTOG and GTSE1 at endogenous levels (see Table 1).
Defining mitotic clathrin, TACC3, chTOG and GTSE1 interactions using knocksideways of endogenously tagged proteins. Acute manipulation of a protein localization using knocksideways can be used to uncover interactions in living cells (Hood et al., 2013). To examine mitotic interactions between clathrin, TACC3, chTOG and GTSE1, we set out to relocalize each endogenous protein in mitotic knock-in cells and ask whether this manipulation affects the localization of the other proteins, detected by indirect immunofluorescence (Figure 3). Relocalization of endogenous clathrin (CLTA-FKBP-GFP) caused the removal of TACC3, GTSE1 but not chTOG from the spindle ( Figure 3A). Similarly, relocalization of endogenous GFP-FKBP-TACC3 resulted in removal of clathrin (very small effect) and GTSE1, but not chTOG from the spindle ( Figure  3B). In contrast, relocalization of either chTOG-FKBP-GFP or GTSE1-FKBP-GFP had no effect on the spindle localization of the other three respective proteins ( Figure 3C,D). While these experiments were designed to examine interactions between endogenous proteins, the use of immunofluorescence had two drawbacks. Firstly, antibody specificity was an issue. No effect on chTOG localization was seen after relocalization of each of the four proteins, including chTOG-FKBP-GFP itself; and relocalization of GTSE1-FKBP-GFP was not detected by anti-GTSE1. Secondly, single cells could not be tracked during knocksideways. We next sought to repeat these experiments using a live cell approach. To do this, the knock-in cell lines were transfected with dark mitotrap and either mCherry-CLTA, mCherry-TACC3, chTOG-mCherry or tdTomato-GTSE1. Metaphase cells were imaged live as rapamycin was added ( Figure 4). Relocalization of endogenous clathrin caused the removal of mCherry-TACC3, chTOG-mCherry and tdTomato-GTSE1 from the spindle ( Figure 4A). Similarly, relocalization of endogenous TACC3 also caused the removal of the other three proteins from the spindle but to a lesser extent than with clathrin ( Figure 4B). Again, relocalization (A) Strategy to tag clathrin (CLTA), TACC3, chTOG/CKAP5 and GTSE1 with FKBP and GFP at their endogenous loci. Cas9n D10A nickase was used to target the indicated site and a repair template with FKBP-GFP or GFP-FKBP tag flanked by left and right homology arms (LHA/RHA) was used as indicated. Scale bar, 1000 bp. GFP-positive cells were individually sorted by FACS and validated using a combination of Western blotting (B), imaging (C) and DNA sequencing (not shown).
(B) Western blotting showed negative clones and positive clones that were either homozygous (single band, shifted by ∼30 kDa) or heterozygous (two bands, one at expected size and the other shifted by ∼30 kDa) knock-in cell lines. The tagged and untagged proteins are denoted by filled green and open gray arrowheads, respectively. Clones used in this work are highlighted in bold. HeLa cells may have more than two alleles of the targeted gene. We use the term homozygous to indicate editing of all alleles and heterozygous to indicate that at least one allele was edited and that an unedited allele remained. PCR and DNA sequencing confirmed that: CLTA-FKBP-GFP (clone 5) is homozygous, GFP-FKBP-TACC3 (clone D5) is homozygous, chTOG-FKBP-GFP (clone H5) is heterozygous, and GTSE1-FKBP-GFP (clone B5) is homozygous, (C) Micrographs of each tagged cell line indicating the correct localization of each tagged protein. Scale bar, 10 µm. of either chTOG or GTSE1 had no effect on the spindle localization of the other three respective proteins ( Figure  4C,D). A semi-automated analysis procedure was used to measure induced relocalization of both proteins (see Methods). All movement was from the mitotic spindle to the mitochondria, without significant loss to the cytoplasm, suggesting that the complex is either relocalized en masse or not. 2D arrow plots were therefore used to visualize the results of these experiments ( Figure 4). As previously reported, mCherry-TACC3 expression distorted the localization of the complex prior to knocksideways (Booth et al., 2011;Nixon et al., 2015), enhancing the amount of clathrin, chTOG and GTSE1 on the spindle (Figure 4, note the rightward shift of the starting point in the arrow plots when mCherry-TACC3 was expressed). This likely reflects the importance of TACC3 in loading the complex onto the spindle (Hood et al., 2013).
The expression of other partner proteins mCherry-CLTA, chTOG-mCherry, and tdTomato-GTSE1 had no effect on the localization of the knock-in protein.
The results of both knocksideways approaches are summarized in Supplementary Table 2. Relocalization of either clathrin or TACC3 during metaphase results in removal of the entire TACC3-chTOG-clathrin-GTSE1 complex. The efficiency of this removal is higher with clathrin than TACC3. Relocalization of either chTOG or GTSE1 has no effect on the rest of the complex, suggesting that these proteins are ancillary to TACC3-chTOG-clathrin-GTSE1.

Role of LIDL motifs in recruitment of GTSE1 to the TACC3-chTOG-clathrin complex.
In order to test if GTSE1 is an ancillary complex member, we sought to disrupt its interaction with clathrin and assess whether or not the This strategy can also be used to assess if another protein Y, co-reroutes with X to the mitochondria. Y1 co-reroutes with X, indicating that they form a complex, whereas Y2 does not. (B) Live cell imaging of knocksideways of gene edited cell lines. Indicated tagged cell lines expressing were imaged on a widefield microscope. Stills from a movie where metaphase cells were each treated with rapamycin 200 nM are shown. The post-rapamycin image is ∼15 min after treatment. Scale bar, 10 µm. spindle-binding of these two proteins was interdependent. To examine the effect on the mitotic localization of both proteins, mCherry-tagged GTSE1 constructs were expressed in GTSE1-depleted CLTA-FKBP-GFP cells ( Figure 5). GTSE1 has a previously-mapped clathrin interaction domain (CID, 639-720), containing five clathrin box-like motifs LI[DQ] [LF] (hereafter referred to as LIDL motifs), which was targeted for disruption (Wood et al., 2017;Rondelet et al., 2020). We found that deletion of the entire CID resulted in a reduction in GTSE1 on the spindle. Mutation of LIDL motifs 1-2, 3 or 4-5 to alanines did not result in reduction, but when mutated in combination resulted in a loss of GTSE1 that was similar to deletion of the CID. However, under all conditions the spindle localization of clathrin was unaffected.
To test if the reduction in GTSE1 spindle localization represented a block of recruitment, cells were treated with 0.3 µM MLN8237 to inhibit Aurora A activity and provide a reference for minimal recruitment (Hood et al., 2013;Booth et al., 2011). Spindle localization of both clathrin and GTSE1 WT was abolished by drug treatment ( Figure S3). Again, spindle localization of GTSE1∆1,2,3,4,5 was lower than WT in untreated cells, and was not reduced further by MLN8237 treatment (p = 0.08). These data are consistent with the idea that GTSE1 is recruited to the spindle by clathrin via multiple LIDL motifs in GTSE1 (Rondelet et al., 2020). Moreover they suggest that there is no interdependent spindle localization of clathrin-GTSE1 and that GTSE1 is an ancillary member of the complex. The ability to bind clathrin is necessary for GTSE1 to localize to the spindle, but is it sufficient? To address this question we examined the subcellular localization of a panel of GTSE1 fragments in mitosis or interphase cells ( Figure 6). A GTSE1 fragment comprising the CID containing all five LIDL motifs was unable to bind the mitotic spindle. Progressively adding more N-terminal sequence to the CID eventually yielded a construct that bound the spindle (161-720, Figure 6A-C). This experiment demonstrated that the CID alone is not sufficient for spindle localization. Interphase microtubule-binding was seen for the GTSE1 fragment 161-720 and to a lesser extent for 1-354, 335-720 and 400-720 ( Figure 6A,D). This suggests that the region 161-638 contains one or more regions that can bind microtubules and that these regions, together with the 5XLIDL motifs in the CID are required for spindle localization.

PIK3C2A
is not a component of the TACC3-chTOG-clathrin-GTSE1 complex. We next    investigated whether or not PIK3C2A is a component of the TACC3-chTOG-clathrin-GTSE1 complex, since PIK3C2A has been proposed to bind TACC3 and clathrin, and therefore stabilize the complex (Gulluni et al., 2017). If PIK3C2A binds the complex, we would predict that it should also localize to the mitotic spindle. We imaged GFP-PIK3C2A in live cells and found no evidence for spindle localization ( Figure 7A). The construct localized to clathrin-coated pits suggesting that the GFP-tag had not interfered with its normal localization. We next overexpressed mCherry-TACC3 to concentrate the TACC3-chTOG-clathrin-GTSE1 complex on the spindle and maximize our chances of seeing any GFP-PIK3C2A signal on microtubules, but again, we saw no spindle localization of GFP-PIK3C2A ( Figure 7B).
To further explore any mitotic role for PIK3C2A, we generated a PIK3C2A knockout cell line using CRISPR/Cas9. This generated a clone with a premature stop codon in both alleles, resulting in 1-87 and 1-72 residues, that we termed PIK3C2A null (Supplementary Figure S4C). It was previously shown that PIK3C2A knockout in primary mouse embryo fibroblasts (MEFS) altered their mitotic progression (Gulluni et al., 2017). We analyzed mitotic timings of our PIK3C2A null cell line, compared to parental HeLa cells, and found no differences in mitotic timings (Supplementary Figure S4D).
If PIK3C2A was a scaffold protein for the TACC3-chTOG-clathrin-GTSE1 complex, we would expect some disruption of the spindle localization of clathrin, TACC3, chTOG or GTSE1 in the PIK3C2 null cells. However, immunostaining of parental HeLa and PIK3C2A null cells with antibodies against CHC, TACC3, chTOG and GTSE1, revealed a similar distribution of all four complex members during mitosis ( Figure 7C).
In the original paper, immunostaining of PIK3C2A at the mitotic spindle was shown (Gulluni et al., 2017). We immunostained parental HeLa and the PIK3C2A-null cells with the same anti-PI3KC3A antibody used in the original report and found that there was a signal at the mitotic spindle, but that it was non-specific since it was also detected in the PIK3C2A-null cells ( Figure 7D). We also used RNAi of PIK3C2A in parental and PIK3C2A-null cells to rule out the possibility that the antibody signal resulted from residual expression of PIK3C2A. Again, the spindle fluorescence remained after RNAi indicating that the antibody is non-specific for immunofluorescence. Taken together, our results suggest that PIK3C2A is not a component of the TACC3-chTOG-clathrin-GTSE1 complex.

Discussion
Inducible relocalization is a powerful method to investigate protein-protein interactions in cells and to pinpoint where and when they occur. We generated a number of cell lines to study the interactions between members of the TACC3-chTOG-clathrin-GTSE1 complex on mitotic spindles at metaphase. This approach showed that TACC3 and clathrin are core complex members, while chTOG and GTSE1 are ancillary. Our current picture of this multiprotein complex is outlined in Figure 8.
It had been reported that PIK3C2A is a component of the TACC3-chTOG-clathrin-GTSE1 complex, where it was proposed to act as a scaffold protein binding both TACC3 and clathrin (Gulluni et al., 2017). This proposal was consistent with several observations. First, PIK3C2A was found to interact with clathrin, GTSE1 and TACC3 in a proteomic analysis of immunoprecipitations with each of these three proteins from mitotic lysate (Hubner et al., 2010).
In that study, immunoprecipitation of PIK3C2A brought down clathrin, GTSE1 and components of the membrane trafficking machinery, but notably neither TACC3 nor chTOG co-immunoprecipitated with PIK3C2A. Second, PIK3C2A binds clathrin heavy chain via an N-terminal region which contains a clathrin box-like motif (LLLDD) (Gaidarov et al., 2001). These motifs bind to grooves in the seven-bladed beta-propeller that constitutes the N-terminal domain of clathrin heavy chain (Smith et al., 2017). The N-terminal domain is required for clathrin-TACC3 to localize at the spindle (Royle et al., 2005) and mutations in one of the grooves is sufficient to reduce spindle-binding (Hood et al., 2013). However, while the proposal that PIK3C2A was a component of the complex made sense, we found no evidence to suggest that PIK3C2A was even present on mitotic spindles. GFP-tagged PIK3C2A was found in clathrin-coated vesicles as expected, but was absent from the mitotic spindles of HeLa cells. We also found that the PIK3C2A antibody used in the original study to detect the protein at the spindle gave a false signal that remained after knockout and/or knockdown of PIK3C2A. Finally, a PIK3C2A-null cell line we generated had no mitotic delays, and all members of the TACC3-chTOG-clathrin-GTSE1 complex had normal localization. We conclude that PIK3C2A is not a component of the complex and any mitotic function for this protein is doubtful. PIK3C2A has a well-established role in clathrin-mediated membrane traffic (Posor et al., 2013) and it seems likely that the presence of PIK3C2A among other membrane trafficking factors in the original proteomic work was due to association with a fraction of clathrin that was not associated with the spindle, or erroneous binding during purification (Hubner et al., 2010).
Recent work has shown that GTSE1 contains five conserved LIDL motifs an intrinsically disordered C-terminal region which can bind to the N-terminal domain of clathrin heavy chain (Rondelet et al., 2020). In agreement with this, we found that these motifs are redundant and that mutations which reduce the total number of motifs to below three, impaired spindle-binding significantly. We also found that GTSE1 was an ancillary protein not required for the localization of the complex on microtubules and that inducing its mislocalization did not affect the other complex members, this interpretation is consistent with other work on GTSE1 (Rondelet et al., 2020;Bendre et al., 2016). It is a mystery why mutation of one groove of the N-terminal domain of clathrin heavy chain results in loss of the complex from the spindle, since it appears that this domain recruits GTSE1 to k-fibers but that GTSE1 is not needed for localization (Hood et al., 2013). One explanation is that this groove interacts directly with microtubules and GTSE1 may also bind other sites on clathrin's N-terminal domain. Another is that the GTSE1-clathrin interaction may be important for the formation of the complex, but not for its stability once loaded onto microtubules.
The ancillary nature of GTSE1 and chTOG binding to the complex via association with clathrin and TACC3 respectively is intriguing. Especially since GTSE1 and chTOG each have the ability to bind microtubules themselves (Monte et al., 2000;Spittle et al., 2000). Rondelet et al. proposed that clathrin-TACC3 could be forming a "scaffold" for the recruitment of other factors, such as GTSE1, to the spindle so that they can in turn perform specific functions (Rondelet et al., 2020;Bendre et al., 2016). Our work is consistent with this idea, that clathrin-TACC3 are core to spindle microtubule-binding and that other ancillary factors may be recruited through this complex. In this work we mapped a constitutive microtubule-binding region in GTSE1 to residues 161-638, while chTOG likely binds the microtubule lattice through a region between TOG4 and TOG5. The criterion for binding clathrin-TACC3 at the spindle may include the ability to bind microtubules, which would explain the selectivity for ancillary partners and mean that clathrin adaptors for example are not recruited to the spindle. Our work establishes that, in order to disrupt the TACC3-chTOG-clathrin-GTSE1 complex, agents that target i) the TACC3-clathrin interaction or ii) the interface between TACC3-clathrin and microtubules are required. In the first case, preventing the helix that is formed by phosphorylation of TACC3 on serine 558 from binding to the helical repeat in the ankle region of clathrin heavy chain is predicted to disrupt the complex (Burgess et al., 2018). Second, the microtubule interface needs to be mapped at high resolution using cryo-electron microscopy.
The endogenously tagged cell lines we have developed will be useful for pursuing these questions. Besides fluorescence microscopy and knocksideways, the cells are well suited for visualizing proteins at the ultrastructural level using inducible methodologies such as FerriTagging (Clarke and Royle, 2018).
For GFP-PIK3C2A a ScaI-BstEII fragment from full-length human PIK3C2A in PCR-XL-Topo (IMAGE:8322710) was cloned into pEGFP-C1. The mCherry-tagged GTSE1 constructs were made by PCR amplification of GTSE1 (IMAGE: 4138532) followed by insertion into pmCherry-N1 between SalI-BamHI, and using site-directed mutagenesis to introduce each mutation.

Cell culture. HeLa cells (Health Protection Agency/European
Collection of Authenticated Cell Cultures, #93021013) were maintained in Dulbecco's Modified Eagle's Medium (DMEM) supplemented with 10 % FBS and 100 U ml −1 penicillin/streptomycin in a humidified incubator at 37 • C and 5 % CO2.
Knock-in cell lines were generated by CRISPR/Cas gene editing.
Details of the repair templates are available (see below). Ten days after transfection, single GFP-positive cells were selected by fluorescence-activated cell sorting (FACS), expanded and validated by microscopy, Western blotting, PCR and DNA sequencing. The PIK3C2A knockout cell line was generated by transfecting HeLa cells with pSpCas9(BB)-2A-GFP (pX458, Addgene #48138) into which a single guide (5'-CACCGAGCACAGGTTTATAACAAGC-3') had been cloned.
GFP-positive cells were isolated by FACS and then single cell clones were validated by western blotting and genome sequencing.
Cells were blocked in 3 % BSA in PBS for 30 min.
Cells were incubated for 1 h at RT with primary antibodies as follows: rabbit anti-Tubulin (PA5-19489, Invitrogen), mouse anti-Tubulin (B-5-1-2, Sigma), mouse anti-CHC (X22, CRL-2228 ATCC), rabbit anti-chTOG (34032, QED Biosciences), mouse anti-TACC3 (ab56595, Abcam), mouse anti-GTSE1 (H00051512-B01P, Abnova), rabbit anti-PIK3C2A (22028-1-AP, Proteintech Data Analysis. Analysis of knocksideways movies was done by extracting a pre-and a post-rapamycin multi-channel image from the sequence. An automated procedure in Fiji measured three regions in each of the following areas: spindle, cytoplasm and mitochondria, after registration of the pre and post images. A background measurement and a whole cell fluorescence measurement were also taken. The average value for each region, after background subtraction, was corrected for bleach using the whole cell fluorescence measurement (background-subtracted) for the respective channel. Data were exported as csv and read into IgorPro (WaveMetrics) where a custom-written procedure analyzed the data and generated all the plots. Ternary diagrams of spindle, mitochondria and cytoplasm fluorescence revealed that knocksideways resulted in movement mainly between spindle and mitochondria (Supplementary Figure S5)). Therefore, the fraction of fluorescence at the spindle and mitochondria were used to generate the arrow plots.
For spindle localization analysis of fixed cells, a 31 × 31 pixel (1.4 µm 2 ) ROI was used to measure three regions of the spindle, the cytoplasm and one region outside of the cell as background, using Fiji. Following background-subtraction, the average spindle fluorescence was divided by the cytoplasm fluorescence to give a measure of spindle enrichment. To quantify the MT localization of GTSE1 fragments, a line-scan analysis method adapted from (Hooikaas et al., 2019) was used.
Using an automated procedure in Fiji, average fluorescence intensities from three lines, 1 µm to 3 µm length, along MTs stained for α-tubulin and three adjacent lines (not coincident with MTs) were measured. Following background subtraction, the average fluorescence intensity of the MT line scan was divided by the average fluorescence intensity of the adjacent control line scan to generate a MT enrichment ratio. Analysis was done by an experimenter blind to the conditions of the experiment. All figures were made in Fiji, R or Igor Pro 8 and assembled using Adobe Illustrator.
Data and software availability. All code used in the manuscript and sequences for repair templates is available at https://github.com/quantixed/p053p030.  Arrows show the fraction of spindle and mitochondria fluorescence that is at the spindle (i.e. 1 = completely spindle-localized, 0 = mitochondria-localized), for both channels, moving from pre to post rapamycin localization. Black arrows represent individual cells, the orange arrow is the mean. n = 7-12 cells per condition.

Figure S3. Comparison of GTSE1 LIDL motif ablation with the effect of Aurora-A inhibition on spindle localization of clathrin and GTSE1.
Representative widefield micrographs of CLTA-FKBP-GFP cells at metaphase to show the spindle localization of clathrin (A), or GTSE1-mCherry construct (WT or ∆1,2,3,4,5, red) and clathrin (B). Cells were treated with control (GL2, Ctrl) or GTSE1 siRNA and Aurora-A kinase was inhibited with MLN8237 (0.3 µM, 40 min) as indicated. Cells were stained for tubulin (red in A, not shown in merge in B) and DNA (blue). A GFP-boost antibody was used to enhance the signal of CLTA-FKBP-GFP (green). Scale bar, 10 µm. (C) Quantification of clathrin and GTSE1 spindle recruitment. Each dot represents a single cell, n = 10-15 cells per condition. The large dot and error bars show the mean and the mean ±SD, respectively. Analysis of variance (ANOVA) with Tukey's post-hoc test was used to compare the means between each group, using the untreated cells + siRNA Ctrl (clathrin) and untreated cells + WT GTSE1 (GTSE1) for comparison. The p-value level is shown compared to WT: ***, p < 0.001; **, p < 0.01; NS, p > 0.05.