Kif2 microtubule depolymerase is required for unequal cell division and localizes to a subdomain of cortical endoplasmic reticulum

Unequal cell division (UCD) is a fundamental process responsible for creating sibling cell size asymmetry; however, how microtubules are specifically depolymerized on one aster of the mitotic spindle creating a smaller sibling cell during UCD has remained elusive. Using invertebrate chordate embryos (ascidian) that possess a large cortical structure (CAB) that causes UCD, we identified a microtubule depolymerase (Kif2) involved in creating cell size asymmetry. Kif2 localizes to the cortical subdomain of endoplasmic reticulum in the CAB. During three successive UCDs, Kif2 protein accumulates at the CAB during interphase and is delocalized from the CAB in mid mitosis. Rapid imaging of microtubule dynamics at the cortex revealed that microtubules do not penetrate the CAB during interphase. Inhibition of Kif2 function prevents the development of mitotic aster asymmetry and centrosome movement towards the CAB thereby blocking UCD, whereas locally increasing microtubule depolymerization causes exaggerated asymmetric spindle positioning. This study provides insights into the fundamental process of UCD and for the first time shows that a microtubule depolymerase is localized to a cortical site controlling UCD.

size asymmetry; however, how microtubules are specifically depolymerized on one aster of the mitotic spindle creating a smaller sibling cell during UCD has remained elusive. Using invertebrate chordate embryos (ascidian) that possess a large cortical structure (CAB) that causes UCD, we identified a microtubule depolymerase (Kif2) involved in creating cell size asymmetry. Kif2 localizes to the cortical subdomain of endoplasmic reticulum in the CAB.
During three successive UCDs, Kif2 protein accumulates at the CAB during interphase and is delocalized from the CAB in mid mitosis. Rapid imaging of microtubule dynamics at the cortex revealed that microtubules do not penetrate the CAB during interphase. Inhibition of Kif2 function prevents the development of mitotic aster asymmetry and centrosome movement towards the CAB thereby blocking UCD, whereas locally increasing microtubule depolymerization causes exaggerated asymmetric spindle positioning. This study provides insights into the fundamental process of UCD and for the first time shows that a microtubule depolymerase is localized to a cortical site controlling UCD.  (Wessel et al., 2014) and C. elegans 1-cell zygotes (Galli and van den Heuvel, 2008). Such UCD relies on the development of unbalanced forces that has two components: cortical pulling force acting on astral microtubule plus ends (Grill et al., 2001) and depolymerization of microtubule plus ends as they encounter the cortex (Kozlowski et al., 2007;Labbé et al., 2003) which create unbalanced forces to position the mitotic spindle asymmetrically. These two processes acting together overcome the forces that cause mitotic spindles to move to the center of the cell, a process that senses and integrates force over the length of microtubules (Minc et al., 2011). However, one key piece missing from this model of UCD is the identity of the protein(s) that cause astral microtubule plus end depolymerization at the cortex, which is important not only for UCD but also for the mitotic spindle centering mechanism based on astral microtubule length that operates during symmetric cell division.
Microtubule plus ends can be induced to depolymerize via different mechanisms. In vitro experiments indicate that dynein can cause catastrophe of microtubule plus ends (Laan et al., 2012), raising the possibility that in intact cells dynein couples pulling with depolymerization. A different mechanism regulates microtubule plus end depolymerization in developing mammalian neurites which is dependent on the cortically-localized microtubule depolymerase Kif2A (Homma et al., 2003). Kif2A is a member of the kinesin-13 family of microtubule depolymerases (Lawrence et al., 2004) which includes MCAK/Kif2C that causes microtubule plus end depolymerization at kinetochores during anaphase (Wordeman et al., 2007). However, in cells that divide unequally it is still not known what causes astral microtubule plus end depolymerization at the cortex, and the only protein known to limit microtubule growth at the cortex during UCD in one cell C.elegans embryos does not affect astral microtubules (O'Rourke et al., 2010).
C.elegans embryos have provided a wealth of knowledge about the cortical pulling forces that act upon astral microtubules of the spindle during UCD. For example, following fertilization and symmetry breaking in C. elegans, the Par polarity complexes are partitioned to distinct cortical subdomains (Gönczy, 2008). Anterior Par3/Par6/aPKC (PKC-3) phosphorylates LIN-5 (NuMA) at the anterior cortex inhibiting the cortical anterior spindle pulling forces (Galli et al., 2011) ,while NuMAs binding partner GPR-1/2 (Pins/LGN) becomes enriched at the posterior cortex during mitosis (Gotta et al., 2003). The Dynactin/Dynein complex protein DRBY-1 interacts with Pins and NuMA thus providing a cortical link to microtubules (Couwenbergs et al., 2007). During metaphase and anaphase the mitotic spindle is pulled towards the posterior cortex causing UCD (Grill et al., 2001). Late in mitosis the posterior centrosome has changed from a spherical shape to a flattened and discshaped structure (Severson and Bowerman, 2003). Symmetric cell divisions in somatic cells also depend upon cortical dynein to center the mitotic spindle (Collins et al., 2012;Kern et al., 2016;Kiyomitsu and Cheeseman, 2013). In addition to dynein pulling forces, it has been shown that astral microtubule depolymerization also plays a role in posterior spindle displacement. For example, pulling forces are lacking when microtubules are stabilized by taxol, while in embryos carrying a temperature-sensitive mutation in a β-tubulin gene the posterior displacement distance of the spindle is enhanced (Nguyen-Ngoc et al., 2007). Based on these and other data a dual force-generation mechanism has been proposed that relies on microtubule pulling forces (dynein-dependent) combined with microtubule depolymerization (Kozlowski et al., 2007;Nguyen-Ngoc et al., 2007). Thus in C.elegans, although cortical microtubule depolymerization is thought to be part of the mechanism for posterior spindle displacement, the mechanism regulating cortical microtubule depolymerization is not known (Gönczy, 2008).
Many embryos provide more extreme examples of UCD whereby the two asters of the mitotic spindle become highly asymmetric in size and shape with the smaller of the two asters being inherited by the smaller of the two daughter cells. Similar to the flattened posterior centrosome in C.elegans one-cell embryos (Severson and Bowerman, 2003), such mitotic aster asymmetry has also been observed during UCD in spiralian (Lambert, 2010;Rabinowitz and Lambert, 2010), echinoderm (Holy and Schatten, 1991;Schroeder, 1987), and ascidian (Hibino et al., 1998;Prodon et al., 2010) embryos. Mitotic aster asymmetry commonly occurs at the third cleavage in spiralian embryos with the smaller aster associating with an animal cortical domain at the 4-cell stage leading to UCD (Rabinowitz and Lambert, 2010).
A similar phenomenon has been documented for sea urchin embryos where one aster in each future micromere associates with a cortical domain situated at the vegetal pole of the 8-cell stage embryo creating aster asymmetry during UCD (Schroeder, 1987;Holy and Schatten, 1991). The clearest example of a cortical domain that causes aster asymmetry and UCD comes from ascidian embryos. Here a large cortical structure termed the CAB (centrosomeattracting body) causes three successive rounds of UCD accompanied by the CAB-proximal aster becoming smaller (Hibino et al., 1998;Nishikata et al., 1999;Prodon et al., 2010).
Although the Par complex (aPKC, Par3 and Par6) is localized to the CAB (Patalano et al., 2006), we do not currently know how the CAB affects microtubule dynamics leading to aster asymmetry. Due the large dimensions of the CAB (circa 20µm at the 8-cell stage) we wondered whether it would be possible to identify proteins involved in regulating microtubule dynamics at the cortex via live cell imaging during UCD in the early ascidian embryo.
We have developed the optically transparent ascidian species Phallusia mammillata as a system to perform live cell imaging to study microtubule dynamics during UCD (McDougall et al., 2015;Prodon et al., 2010). By analyzing microtubule dynamics at the cortex we discovered that a microtubule depolymerase (Kif2) localizes to the cortical CAB in a cell-cycle dependent manner. Through live cell imaging of Kif2::Venus/mCherry/Tomato and immunofluorescence we demonstrate that exogenous and endogenous Kif2 localizes to the cortical CAB. In particular, Kif2 accumulates on a domain of cortical endoplasmic reticulum (cER) concentrated at the CAB during interphase and leaves the CAB cER during mid mitosis when CAB proximal microtubules become short. In addition, we show that microtubules do not penetrate the cortical CAB during interphase. Finally, we found that inhibiting endogenous Kif2 protein function prevents the establishment of mitotic aster asymmetry and UCD, and conversely that increasing depolymerization of microtubules near the CAB causes the CAB proximal spindle pole to move nearer to the CAB.

Unequal cell division and the CAB
The early embryo of the European ascidian Phallusia mammillata is favorable for livecell imaging and functional studies because its cells are transparent (see Movie S1) and  Figure 1A). During these three rounds of UCD, astral microtubules emanating from the centrosome that polymerize towards the CAB and midline are shorter than those microtubules that do not polymerize towards the CAB or microtubules originating from the more distant centrosome as the mitotic spindle approaches the CAB and midline (Figure 1 A). A smaller and flattened aster thus forms nearest the CAB and midline during each round of UCD ( Figure 1 and Movies S2, S3 and S4 with a 3D rendering of an 8cell stage embryo shown in Movie S5). In Phallusia embryos both centrosomes appeared similar for -tubulin staining (sFig. 1) indicating that aster size is not a function of -tubulin loss from one centrosome as has been observed in leech zygotes (Ren and Weisblat, 2006).
The CAB is a multilayer structure and can be visualized in several ways (Paix et al., 2011a;Patalano et al., 2006). Because the CAB creates a protrusion it can be visualized with plasma membrane markers such as PH::Tom (Figure 1 B and Movie S6). Also, because the CAB is rich in cortical ER and excludes mitochondria it can be visualized by specific lipophilic dyes that label either the mitochondria or cER in the CAB (Figure 1 B). Finally, antibodies to aPKC label the cortical surface of the CAB but do not label the cER which appears as a dark zone surrounded by mitochondria labelled with anti mitochondrial antibody-NN18 (Figure 1 B).

Characterization of the microtubule depolymerase Kif2 in ascidian embryos
In order to understand how the CAB may be involved in creating aster asymmetry we searched for CAB-resident proteins by screening likely candidates by either probing with antibody for immunofluorescence or localization of expressed fluorescently tagged proteins.
We identified a number of proteins including Kif2, a member of the kinesin-13 family of proteins that instead of possessing motor activity displays microtubule depolymerization activity (Hirokawa and Takemura, 2004). Vertebrates contain three members of the Kif2/kinesin-13 family: Kif2a, Kif2b and MCAK (Kif2c) (Hirokawa and Tanaka, 2015).
There is only one member of the Kif2 family in the ascidian (P. mammillata: PmKif2 and C.
intestinalis: CiKif2) and other non-vertebrate deuterostomes (sFig 2), suggesting that the vertebrate family of proteins evolved from a non-vertebrate deuterostome Kif2a/b/c (henceforth Kif2). Consistent with this ascidian Kif2 also localizes to spindle poles (like

Kif2 cycles onto and off of the CAB
We noticed that Kif2 protein appeared to accumulate on the CAB during interphase and leave the CAB during mitosis. In order to confirm that Kif2 cycled onto and off of the CAB we performed live cell ratiometric imaging of Kif2::Tom levels relative to Plk1::Ve or other fluorescent constructs (Par6::Venus, H2B::GFP or PH::GFP). Our analysis revealed that Kif2 protein accumulates at the CAB during interphase peaking at early mitosis and that Kif2 protein begins to leave the CAB during prometaphase with levels reaching a minimum at late anaphase/cytokinesis onset ( Figure 3 A and Movie S7). Careful examination of endogenous Kif2 levels at the CAB in either Phallusia or Ciona embryos revealed a similar dynamic delocalization of Kif2 during mitosis, whereby endogenous KIF2 protein was delocalized from the CAB by late anaphase (Figure 3 B and sFig 4). Other CAB-resident proteins such as PEM1 or aPKC do not leave the CAB during mitosis (Figure 3 B). In order to determine whether loss of signal at the CAB was due to delocalization or degradation we used the photo-convertible construct Kif2::Dendra to follow a specific pool of Kif2 protein in vivo. UV illumination converts the green Kif2::Dendra fusion protein into a red one. UV illumination of Kif2::Dendra in a region within one CAB caused just one CAB to become red (Figure 3 C). Since Kif2 is also a kinetochore localized protein, we reasoned if we could detect red Kif2::Dendra on chromosomes that would indicate that the red version of Kif2 protein left the CAB following photo conversion to become captured by adjacent chromosomes (it is important to note that this does not rule-out destruction but does show that red Kif2 protein leaves the CAB).

Kif2 localizes to cortical ER in the CAB
As noted previously, the CAB is a multilayered structure, comprised of a thick layer of cortical endoplasmic reticulum (cER) protruding into the cytoplasm which adheres to a specialized region of actin-rich cortex (Patalano et al., 2006). By co-staining for mitochondria which surround and outline the cortical ER mass, and for pMNK, a cER resident protein (Paix et al., 2011b), the surface and deeper cER can be visualized (Figure 4 B). Unlike aPKC protein which is contiguous with the actin layer, Kif2 protein occupies the thicker cER layer ( Figure 2). To determine the precise localization of Kif2 protein in the CAB we prepared isolated cortices. By sticking 8-cell stage embryos to coverslips followed by shearing using an isotonic buffer, the cortex and its associated cER is retained on the cover slip (Figure 4 A). Probing these cortical preparations with anti-Kif2 revealed a concentration of Kif2 in the cER of the CAB (Figure 4 A, and inset for higher definition). In live embryos we were able to observe dynamic finger-like projections from the surface of the CAB labelled with Kif2::Tom (Figure 4 C), consistent with the notion that Kif2 was localized to the tubes of cER in the CAB.

Microtubule dynamics at the CAB
To visualize the rapid dynamics of microtubule at the cortex embryos have to be immobilized close to a flat surface, here provided by the coverslip. In one cell C.elegans zygotes this permitted the accurate determination of plus end dynamics leading to the development of the "touch and pull" mechanism of ACD (Kozlowski et al., 2007). In ascidians, measurement of microtubule plus end dynamics in a cortical slice containing the CAB and non-CAB cortex is complicated by both the movement and the geometry of the embryo: since the CAB is an apical cortical structure close to the midline it invariably curves away from the coverslip. In order to overcome these problems we dissociated blastomeres with calcium-free seawater and used protamine-coated coverslips to immobilize them ( Treating embryos in mitosis with taxol also prevented the development of aster asymmetry ( Figure 6 Aii). These data suggest that aster asymmetry and asymmetric spindle positioning require the activity of Kif2 and microtubule depolymerization.
If aster reduction facilitates the migration of one spindle pole towards the cortex, we reasoned that increasing microtubule depolymerization near the CAB would enhance spindle pole movement towards the CAB, since microtubules would become shorter. We employed a method to depolymerize microtubules locally by using a micro-pipette as a spatially-confined source of nocodazole (see materials and methods) ( Figure 6 Bi and Movie S10 and S11).
Since it was not possible to perform these experiments on our confocal microscope we used an epifluorescence system. We bathed embryos in Cell Mask orange which labels both the plasma membrane and vesicles that accumulate around spindle poles during mitosis (McDougall et al., 2015). This was more convenient that using Ens::3GFP and also gave better visualization of both spindle poles in each blastomere. However, to confirm that these nocodazole micropipettes caused localized microtubule depolymerization we imaged microtubules in zygotes (Figure 6 B iv). We applied different diameter pipettes with the same nocodazole concentration near the CAB in one blastomere at the 8-cell stage and measured the effect on the position of the proximal spindle pole. Relatively small diameter nocodazole pipettes caused the proximal spindle pole to migrate closer toward the CAB and midline than normally observed ( Figure 6 Bi and Movie S10). Importantly, the blastomeres far from the source of nocodazole were unaffected and divided normally ( Figure 6 Bi and Movie S10). By using a larger bore pipette we could affect both CAB-containing blastomeres so that both CAB-proximal spindle poles moved closer to the CAB and midline while blastomeres farther from the pipette behaved normally and divided (Figure 6 Bii). Control DMSO-containing pipettes had absolutely no effect on spindle positioning (Figure 6 Biii). To depolymerize microtubules via a second method we used photo-activation of caged Combretastatin while imaging microtubules with Ens::3GFP (Figure 6 C). Activation of caged Combretastatin during mitosis caused astral microtubule depolymerization while leaving spindle microtubules intact. As noted with nocodazole pipettes, the whole spindle migrated even closer towards the CAB (Figure 6 C). These data revealed that reducing aster size could augment the asymmetric positioning of the spindle, supporting the hypothesis that astral microtubule length sets the distance of the spindle pole to the CAB.

DISCUSSION
Here we present evidence for the localization and function of a microtubule depolymerase (Kif2) at a cortical site that causes UCD. Precise control of microtubule dynamics at the cortex is a fundamental cellular mechanism that is involved in diverse  (Schroeder, 1987;Holy and Schatten, 1991) at the vegetal pole where aPKC is absent (Prulière et al., 2011). Again the mechanism underlying this UCD is not known in these embryos, however, observations indicate that one centrosome becomes disc-shaped and closely apposed to the cortex before UCD (Schroeder, 1987;Holy and Schatten, 1991). In ascidians three successive rounds of UCD occur starting at the 8 to 16-cell stage, resulting in the formation of two small blastomeres at the 64-cell stage that are germ cell precursors (Hibino et al., 1998). We showed previously that the CAB causes one pole of the mitotic spindle to approach the CAB during prometaphase through anaphase which is accompanied by the shrinking of the CAB-proximal aster (Prodon et al., 2010).
One central unresolved question therefore is how specialized cortical sites affect astral microtubules leading to the development of aster asymmetry. Here we show that in ascidian embryos the microtubule depolymerase Kif2 plays a key role in promoting aster asymmetry which in turn is required for UCD. Kif2 is concentrated on a subdomain of cortical endoplasmic reticulum (cER) found at the centrosome-attracting body or CAB during interphase and leaves the CAB during mitosis (Figure 4). We propose that Kif2 affects local microtubule dynamics at the cortex in the vicinity of its location, causing depolymerization of the nearest microtubules during mitosis thus leading to the development of aster asymmetry.
By artificially increasing the amount of local depolymerization of astral microtubules in the proximity of the CAB we found that the spindle pole moved even closer to the CAB ( Figure   6).
Our results have led us to the conclusion that polymerization of astral microtubules opposes the pulling forces that likely displace the mitotic spindle towards the CAB. We therefore propose that the diffusion of Kif2 from the CAB in mitosis causes the local depolymerization of those astral microtubules nearest the CAB thus facilitating the eccentric positioning of the mitotic spindle near the CAB cortex.
The mechanism we have discovered here in a chordate deuterostome embryo extends our understanding of UCD, which so far has been heavily studied in two protostomes,

Origin of the animals
Phallusia mammillata were collected at Sète (Etang de Tau, Mediterranean coast, France) and Ciona intestinalis at Roscoff. Ascidian gamete collection, dechorionation, fertilization and embryo cultures were as described previously (McDougall et al., 2014).

Antibodies, fixation and reagents
Embryos were fixed in -20° methanol containing 5uM EGTA and immunolabelled as

Microinjection and imaging
Microinjection was performed as previously described (McDougall et al., 2014).
Briefly, dechorionated oocytes were mounted in glass wedges and injected with mRNA (1-2 µg/µl pipette concentration/ ~2% injection volume) using a high pressure system (Narishige These prepared microinjection needles were stored in humid chambers and used the same day.
The nocodazole/agarose needles were advanced towards the 8-cell stage embryos and placed on the surface of one B4.1 blastomere near the CAB starting at nuclear envelope breakdown.
Bright-field and fluorescence images were acquired every 10 seconds using MetaMorph software package.

Synthesis of RNAs
We used the Gateway system (Invitrogen) to prepare N and C terminal fusion constructs using pSPE3::Venus (a gift from P. Lemaire), pSPE3::Rfp1, pSPE3::Cherry, pSPE3::tomato for all constructs except PH::GFP which was cloned into pRN3. For construct details please refer to our previous methods publication (McDougall et al., 2015) All synthetic mRNAs were transcribed and capped with mMessage mMachine kit (Ambion).

Bioinformatics
We created a database of animal protein sequences derived from the complete genomes       In order to follow microtubule dynamics at the CAB unfertilized eggs were microinjected with mRNA for Ens::3GFP and Kif2::Tom, fertilized and then transferred to calcium-free seawater at the 8-cell stage to dissociate the 8 blastomeres. Once the dissociated blastomeres had divided, pairs of B5.2 blastomeres were placed on coverslips that had been treated with protamine so that the blastomeres adhered. Next fast confocal imaging (0.8 sec/image) of a 1µm thick optical section just above the coverslip revealed that microtubules were absent from the CAB (red    . With the exception of the Phallusia sequence, the first two letters of leaf names represent the species of the sequence (see methods for list). These are followed by identifiers retrievable from the NCBI protein sequence database. Bootstrap support is show above branches. Although within the gene groups the species phylogeny is generally not well resolved, there is high support for lineage specific duplications of Kif2 in Drosophila and Humans, with the majority of invertebrate taxa having only a single representative, strongly suggesting the state in the ancestral animal was a single gene. Ciona and Phallusia data are highlighted by the pink box.

Figure S3. Microtubule density is reduced in cells overexpressing Kif2.
Eggs injected with mRNAs encoding Kif2 (without a fluorescent tag) and Ens::3GFP or only Ens::3GFP (top) were fertilized then treated with cytocholasin B to prevent cell division.
Microtubule density was greatly reduced in the presence of excess Kif2. Scale bar = 20µm.

Figure S4. Kif2 protein is delocalized from the CAB in Ciona intestinalis during mitosis.
Immunolabelling with anti-Kif2 of 16-cell stage Ciona embryos. Kif2 labels the CAB during later interphase/early prophase and is absent at the CAB during anaphase. Scale bar = 30 µm.
Graph showing percentage of embryos (total n=94) with CAB labelled at different stages of the cell cycle.