Loss of MYO5B expression deregulates late endosome size which hinders mitotic spindle orientation

Recycling endosomes regulate plasma membrane recycling. Recently, recycling endosome–associated proteins have been implicated in the positioning and orientation of the mitotic spindle and cytokinesis. Loss of MYO5B, encoding the recycling endosome–associated myosin Vb, is associated with tumor development and tissue architecture defects in the gastrointestinal tract. Whether loss of MYO5B expression affects mitosis is not known. Here, we demonstrate that loss of MYO5B expression delayed cytokinesis, perturbed mitotic spindle orientation, led to the misorientation of the plane of cell division during the course of mitosis, and resulted in the delamination of epithelial cells. Remarkably, the effects on spindle orientation, but not cytokinesis, were a direct consequence of physical hindrance by giant late endosomes, which were formed in a chloride channel–sensitive manner concomitant with a redistribution of chloride channels from the cell periphery to late endosomes upon loss of MYO5B. Rab7 availability was identified as a limiting factor for the development of giant late endosomes. In accordance, increasing rab7 availability corrected mitotic spindle misorientation and cell delamination in cells lacking MYO5B expression. In conclusion, we identified a novel role for MYO5B in the regulation of late endosome size control and identify the inability to control late endosome size as an unexpected novel mechanism underlying defects in cell division orientation and epithelial architecture.


Introduction
Recycling endosomes, strategically positioned at the intersection of endocytic and biosynthetic trafficking pathways, are membrane compartments with versatile functions [1,2] endosomes are important for, among others, epithelial cell polarity development, cell-cell adhesion, and cell fate determination [2]. Mutations in recycling endosome-associated and MYO5B-encoded myosin Vb protein can cause progressive familial intrahepatic cholestasis [3] and/or microvillus inclusion disease (MVID) [4][5][6]. MVID phenotypes are the result of defective plasma membrane recycling in intestinal epithelial cells and intestinal tissue architecture defects. Of interest, loss of MYO5B expression is also associated with gastrointestinal cancer [7,8] and is a strong prognostic factor for colorectal cancer recurrence [9]. Recycling endosomes and therewith associated proteins, notably the small GTPase rab11a, have also been implicated in mitotic cell division processes including cytokinesis [10,11] and, more recently, the organization and orientation of the mitotic spindle apparatus [12][13][14]. Mitotic spindle orientation plays a key role in the development and maintenance of epithelial tissue architecture and may function in tumor suppression [15][16][17][18]. Defects in mitotic spindle and cell division orientation can cause the mispositioning of cells out of the epithelial layer known as cell delamination [19]. Delamination can cause apoptosis or loss of epithelial cell polarity and has been associated with tumorigenesis [18]. Notably, although myosin Vb is a recycling endosome-associated protein and implicated in cell polarity, cancer, and tissue architecture defects, the role of myosin Vb in mitotic cell division has not been addressed.
In this study, we have investigated mitotic cell division following the loss of MYO5B expression in intestinal epithelial cells. It is shown that the loss of MYO5B resulted in delayed cytokinesis, perturbed mitotic spindle orientation, and led to the misorientation of the plane of cell division during the course of mitosis and resulted in the delamination of epithelial cells. Mechanistically, we show that loss of MYO5B expression, in addition to its effects on plasma membrane homeostasis, also affected late endosome homeostasis and size. Surprisingly, we found that the presence of giant late endosomes that formed upon the loss of MYO5B was wholly responsible for mitotic spindle misorientation. These findings add an unexpected new dimension to the role of membrane trafficking proteins in processes that ensure proper mitotic progression and cell division.

Loss of MYO5B causes mitotic spindle orientation defects and cytokinesis delay
In order to investigate the effects of loss of MYO5B on mitotic spindle orientation, we generated MYO5B knockout human intestinal epithelial Caco2 cells using CRISPR-Cas9 technology. Knockout of the MYO5B gene was verified by the absence of the encoded myosin Vb protein on Western blot (see S1A Fig) and supported by sequencing of the MYO5B gene, which revealed the introduction of a premature termination codon in exon 3 (see S1B Fig). MYO5B knockout cells (hereafter referred to as Caco2 MYO5B−/− ) and MYO5B wild-type (WT) cells (hereafter referred to as Caco2 WT ) were cultured on coverslips and transfected with mCherrytagged Histone2B to visualize chromosomes and GFP-tagged β-tubulin or, alternatively, labeled with antibodies against these proteins in order to visualize the mitotic spindle apparatus by laser scanning confocal fluorescence microscopy. The orientation of the mitotic spindle relative to the substratum (x-z direction) in cell monolayers was determined in 3D-reconstructed confocal images(x-y direction) of fixed cells by measuring the angle (β) between a line drawn through the spindle poles and a line parallel to the substratum (Fig 1A). The results show that metaphase Caco2 WT cells showed an average β-angle of 6˚relative to the substratum, and a β-angle beyond 20˚was rarely observed (Fig 1B). In contrast, metaphase Caco2 MYO5B−/− cells showed an average β-angle of 12˚and a significantly higher percentage of cells displayed β-angles beyond 20˚and up to 40˚ (Fig 1A, Fig 1B). This increase in the frequency of tilted The β-angle represents the angle between the spindle axis and the substratum in the confocal x-z mitotic spindles in metaphase Caco2 MYO5B−/− cells persisted, albeit to a lesser extent, through subsequent anaphase (Fig 1C, Fig 1D). Aberrant spindle tilting can result in the delamination of epithelial cells from the monolayer and loss of epithelial polarity of delaminated cells [15][16][17]19,20]. In accordance, in comparison to Caco2 WT cells, Caco2 MYO5B−/− cell cultures showed an increased percentage of cells that were placed on top of the monolayer (Fig 1E, Fig 1F) and which had lost basolateral surface polarity as evidenced by the nonpolarized distribution of the basolateral resident Na + /K + -ATPase ( Fig 1G). Similar results were obtained when the cells were plated as a monolayer on semipermeable Transwell filter supports (see S2 Fig).
In addition to the tilted spindle phenotype, Caco2 MYO5B−/− cells showed a higher cytokinesis index when compared with Caco2 WT cells, as evidenced by the presence of a β-tubulin-positive cytokinesis bridge (Fig 2A). Live cell imaging of mCherry-H2B/GFP-βtubulin-expressing cells revealed that cytokinesis was delayed in Caco2 MYO5B−/− cells (Fig 2B, see S1 Video, S2 Video). Thus, measuring the time that passed from the first appearance until the disappearance of the GFP-β-tubulin-positive cytoplasmic bridge showed an on average 2-fold increase (from 40 to 90 min in Caco2 WT and Caco2 MYO5B−/− cells, respectively) in cytokinesis duration (Fig 2C).
In conclusion, mitotic spindle orientation defects, and cytokinesis delays were observed in Caco2 cells lacking MYO5B expression.

Concurrence of mitotic spindle orientation defects and the presence of large vacuoles in MYO5B-depleted cells
Upon closer microscopical inspection of cells with tilted mitotic spindles, we noticed that virtually all Caco2 MYO5B−/− cells with tilted spindles contained one or more large (i.e., Ø >1 μm) vacuoles ( Fig 3A, arrows and dotted white circles [note that this is the same image as in Fig  1A]). Subsequent comparison of the β-angles in Caco2 MYO5B−/− cells with a vacuole or without a vacuole revealed that the increase in spindle tilting during metaphase and anaphase was restricted to those Caco2 MYO5B−/− cells that displayed a vacuole (Fig 3B, Fig 3C). The absence of a spindle orientation defect in the small intestine of Myo5b knockout mice (see S3 Fig), which do form vacuoles but in villus enterocytes and not in the proliferative crypt cells (see below), supports that the mere loss of Myo5b expression (i.e., without a vacuole present) did not cause the spindle misorientation phenotype.
We then subjected Caco2 MYO5B−/− cells to live cell fluorescence imaging in the x-y plane to investigate the effect of the vacuole on spindle orientation dynamics in real time. In Caco2-MYO5B−/− cells without a vacuole, spindle rotation dynamics were minimal during metaphase ( Fig 4A, see S3 Video). Thus, the average angle (α) between a line drawn through the spindle axis at the onset of metaphase and the end of metaphase (i.e., the onset of anaphase) was approximately 10˚. By contrast, in Caco2 MYO5B−/− cells that contained a vacuole, the average α-angle was approximately 25˚ (Fig 4B). Reconstructed movies of images captured serially in time showed a dynamic reorientation of the mitotic spindle relative to the position of the dimension. (B) The β-angle was quantified in metaphase. (Left side graph) Each dot indicates one cell's β-angle. (Right side graph) The statistical analysis of the mean for each experiment. n � 15 mitotic cells/experiment were analyzed for N = 3 independent experiments. Values for each data point can be found in S1 Data. (C) Caco2 WT and Caco2 MYO5B−/− cells in anatelophase were fixed and stained as indicated. (D) The quantification of β-angle in anatelophase. n � 9 mitotic cells/experiment were analyzed for N = 3 independent experiments. Values for each data point can be found in S1 Data. (E-F) The presence of cells not contacting the substratum was indicated by asterisks in nuclei (E). The percentage of nonsubstrate-contacting cells was quantified. Values for each data point can be found in S1 Data (F). (G) Na + /K + ATPase staining shows basolateral localization in substrate contacting cells but not in multilayered cells. N = 3 independent experiments. t test, � p < 0.05, �� p < 0.01, ��� p < 0.001. Error bars indicate ± SEM (dot graph) or + SD (bar graph). Scale bars: 5 μm.   (Fig 4A; see S4 Video). Quantitative analysis of still images taken from Caco2 MYO5B−/− cells that displayed increased spindle rotation dynamics (i.e., an α-angle of more than 25˚; c.f., Fig 4B) revealed an increase in the distance (d) measured between the vacuole rim and the spindle pole (Fig 4C) during the course of metaphase (Fig 4D and 4E, see S5 Video), supporting that the orientation of the mitotic spindle changed relative to the position of the vacuole. Thus, the presence of the vacuole influenced mitotic spindle orientation.
In addition to the mitotic spindle tilting, the presence of a large vacuole was also correlated with an increased frequency of 2 other mitotic complications: (1) a modest increase (from approximately 3% to approximately 8% of WT and knockout cells, respectively) in metaphase arrest frequency (see S4A- Thus, the mitotic spindle tilting defect and the increased incidence of metaphase arrest and chromosome segregation difficulties-but not the delay in cytokinesis-as observed in MYO5B-depleted cells was strongly correlated to the presence of large vacuoles.

Loss of MYO5B expression causes the formation of giant late endosomes
We then focused on the identity of the vacuoles. We found that Caco2 MYO5B−/− cells, as such, displayed significantly more large vacuoles in the cytoplasm when compared with Caco2 WT cells (Fig 5A), and this was observed in multiple MYO5B knockout clones ( Fig 5B). Large vacuoles were observed in approximately 30% of all fixed cells ( Fig 5A). Live cell imaging, however, revealed that <20% of the Caco2 WT but >80% of the Caco2 MYO5B−/− cells formed a vacuole (Ø > 1 μm) for at least 30 min during the 24 h time course of the experiments (Fig 5C). Electron microscopy analysis confirmed the presence of large electron-lucent vacuoles in Caco2 MYO5B−/− cells (Fig 5D), and the presence of intraluminal material suggested resemblance to degradative compartments. Immunolabeling with antibodies against various organelle markers showed that the vacuoles were specifically positive for late endosome markers (i.e., late endosome-associated membrane protein [LAMP]1 and the small GTPase rab7) (  These results indicate that the observed effects were not specific for Caco2 cells. We reasoned that understanding the mechanism that underlie the formation of giant late endosomes may provide tools to interfere with their formation and, hence, allow us to further test the relationship between the presence of the vacuoles and the mitotic spindle orientation defects. The distended appearance of the late endosomes suggested an ion imbalance and resembled the swollen late endosomes as seen in cells infected with Helicobacter pylori. H. pylori is a bacterium that inserts a vacuolating toxin with chloride channel activity into the host cell's endosomal system and in this way causes late endosome ion imbalance and resultant late endosome swelling [21][22][23]. Indeed, inhibition of the vacuolar H + -adenosine triphosphatase (vATPase) proton pump activity with bafilomycin A1 (BafA1) or inhibition of chloride channel activity with either of 3 unrelated compounds (i.e., furosemide, 4,4'-Diisothiocyano-2,2'-stilbenedisulfonic acid [DIDS] and 5-nitro-2-(3-phenylpropyl-amino) benzoic acid [NPPB]) effectively and reversibly abolished the appearance of giant late endosomes in Caco2 MYO5B−/− cells (Fig 6A and 6B). Possibly, late endosome ion imbalance in MYO5Bdepleted cells is the result of inhibited plasma membrane recycling of various ion transporters and their resultant accumulation in late endosomes. Indeed, loss of MYO5B expression is known to inhibit plasma membrane protein recycling and result in the appearance of plasma membrane proteins in late endosomes. Furthermore, in support of this, we observed a redistribution of a chloride channel, anoctamin (ANO)-6 [24,25], from the cell periphery to (giant) late endosomes upon loss of MYO5B expression (Fig 6C and 6D).
Notably, treatment of Caco2 MYO5B−/− cells with BafA1 not only rescued the giant late endosome phenotype but also normalized the tilted spindle orientation to control values (Fig 7A-7C). Treatment of WT cells with BafA1 did not affect spindle orientation (Fig 7D and 7E). Further treatment of Caco2 MYO5B−/− cells with BafA1 abolished cell delamination (Fig 7F and 7G). These results support the relationship between the presence of the vacuoles and the mitotic spindle orientation defects.

Rab7 availability controls the formation of giant late endosomes and mitotic spindle orientation defects in MYO5B-depleted cells
Rab7 is a key regulator of late endosome dynamics [26] and has been reported to correct trafficking defects in lysosomal storage disease when overexpressed [27]. Our data demonstrate that the overexpression of an enhanced green fluorescent protein (eGFP)-tagged WT (rab7-WT) or eGFP-tagged dominant-negative (DN) rab7 mutant (rab7-T22N) in Caco2-MYO5B−/− cells inhibited or exaggerated the formation of giant late endosomes, respectively ( Fig 8A and 8B).
We took advantage of the modifying effects of rab7 to further address the causality between the presence of giant late endosomes and mitotic spindle orientation defects upon the loss of MYO5B expression. We found that the overexpression of rab7-WT, but not of the inactive rab7-T22N mutant, in Caco2 MYO5B−/− cells not only effectively prevented the formation of each data point can be found in S3 Data. t test, �� p < 0.01, ���� p < 0.0001. Error bars indicate ± SEM (dot graph) or + SD (bar graph). Scale bars: 5 μm.
https://doi.org/10.1371/journal.pbio.3000531.g003 giant late endosomes but also normalized the mitotic spindle tilting range relative to the substratum (β-angle) in metaphase and anaphase cells (Fig 9A-9C). Furthermore, the overexpression of rab7-WT, but not of rab7-T22N, prevented the increase in Caco2 MYO5B−/− cell delamination (Fig 9D and 9E). The expression of rab7-WT or rab7-T22N in Caco2 MYO5B−/− cells did not affect the observed delay in cytokinesis (see S11 Fig), which is in agreement with the nonsignificant correlation between the presence of giant late endosomes and cytokinesis duration. Together these results indicate a causal relationship between the presence of giant late endosomes, spindle orientation defects, and epithelial cell delamination, but not cytokinesis delay, in cells that have lost MYO5B expression.

Wortmannin-induced giant endosomes correlate with mitotic spindle orientation defects
We reasoned that if giant endosomes, rather than the mere loss of MYO5B, was responsible for the mitotic spindle orientation phenotype, the induction of giant endosomes via a different mechanism should also result in the spindle orientation phenotype. To address this issue, we treated WT Caco2 cells with wortmannin (Wm), a fungal metabolite and potent inhibitor of phosphatidylinositol-3 kinases that was previously reported to induce endosomal vacuoles in different cell types [28][29][30]. Treatment of WT Caco2 cells with Wm induced the formation of large endosomal vacuoles with comparable sizes as seen in MYO5B-depleted cells ( Fig 10A). Antibody labeling showed that these vacuoles were positive for markers of both early (EEA1) and late (rab7, LAMP1) endosomes ( Fig 10B). Wm-treated WT Caco2 cells that formed large vacuoles displayed a mitotic spindle tilting phenotype similar to that seen in vacuole-containing MYO5B-depleted cells (Fig 10C-10E). These data thus support a physical occlusion model in which the presence of giant endosomes, rather than other myosin Vb-mediated processes, caused tilting of the mitotic spindle apparatus.

Discussion
In this study, we demonstrate that the presence of giant late endosomes in the cytoplasm of mitotic epithelial cells can influence the orientation of the mitotic spindle apparatus, lead to the misorientation of the plane of cell division during the course of mitosis, and increase the delamination of epithelial cells. This is the first evidence that aberrant endosome size, presumably through physically hindering the alignment of the mitotic spindle with the substratum, can derange cell division orientation and epithelial monolayer organization and, hence, underscores the need for proliferative epithelial cells to restrict and maintain (late) endosome size when entering mitosis.  Large organelles such as the Golgi apparatus are typically fragmented into smaller units prior to mitotic entry, and the inability of cells to fragment their Golgi prevents mitotic entry [31]. Endosomes in mammalian cells are relatively small with typical narrow size distributions and do not show fragmentation during mitosis [32,33]. Except perhaps for the observed metaphase arrest in some cells, the presence of giant late endosomes did not robustly trigger a mitotic checkpoint. This suggests that cells do not actively monitor endosome size in mitosis and, therefore, that the mitotic process is vulnerable to deregulated endosome size.
Giant late endosomes developed as a direct result of loss of MYO5B expression, which was supported in vivo by the presence of aberrantly sized late endosomes in intestinal epithelial cells in Myo5b knockout mice and MVID patients with MYO5B mutations. Disorganized LAMP1-positive compartments were previously also reported in Myo5b knockout (KO) mice by Engevik and colleagues [34] and, using electron microscopy, Vogel and colleagues reported numerous large lysosomes among other vesicular-tubular compartments in MVID patients' enterocytes [35]. MYO5B encodes the recycling endosome-associated myosin Vb protein which, when mutated in MVID patients, causes in villus epithelial cells the redistribution of cell surface proteins to intracellular vesicles among which are late endosomes [4,36,37]. The formation of giant late endosomes in MYO5B-depleted cells was dependent on proton pump and chloride channel activity. Together with their vacuolated morphology at the ultrastructural level and the observed redistribution of chloride channels to late endosomes in MYO5Bdepleted cells, it is conceivable that the giant late endosomes formed as a result of perturbed   Values for each data point in panels A, D, and E can be found in S10 Data. EEA, early endosomal atigen-1; LAMP, late endosome-associated membrane protein; Wm, Wm; WT, wild type. endosome phenotype and for the mitotic spindle orientation defect. It is possible that many cargoes with small but additive effects are responsible.
Our results are particularly significant because previous studies have shown that recycling endosomes and therewith associated rab11a contributed to the regulation of mitotic spindle positioning and orientation [12,13]. The mitotic defects resulting from the loss of rab11a [12] or MYO5B (this study) appear distinct at the mechanistic level because loss of rab11a did not induce giant late endosomes [38][39][40] as observed here in MYO5B-depleted cells, and loss of MYO5B did not cause metaphase delay or affect astral microtubules (this study) as reported in rab11a-inhibited cells [12]. Moreover, spindle orientation defects in MYO5B-depleted cells could be mimicked fully rescued by overexpression of the late endosome-associated rab7. The mechanisms via which rab7 achieves this is not clear, but the finding further supports the requirement of late endosomes to induce the spindle orientation defects upon loss of MYO5B expression. Our in vivo data from Myo5b knockout mice, which developed late endosome vacuoles in their villus cells but not in their proliferative crypt cells, indicate that the mere loss of Myo5b, i.e., without vacuoles present, is not sufficient to induce a spindle orientation defect. In agreement, the observation that the induction of endosomal vacuoles in WT cells by Wm similarly resulted in tilted mitotic spindles and cell delamination support a model in which it was the presence of giant endosomes, rather than other myosin Vb-mediated processes, that caused tilting of the mitotic spindle apparatus Together, the data presented in this study thus identify deregulated late endosomal size as another mechanism that can contribute to aberrant mitotic spindle orientation in response to the loss of a recycling endosome-associated protein. These findings add an unexpected new dimension to the role of membrane trafficking proteins in processes that ensure proper mitotic progression and cell division.
It should be noted that although virtually all mitotic cells that showed increased tilting of their spindle apparatus contained a giant late endosome, not all mitotic cells with a giant late endosome showed increased tilting of their spindle apparatus and multilayering. This apparent discrepancy, i.e., that not all cells with a giant late endosome showed such mitotic phenotype, may be attributed to (i) the size and position of the vacuole relative to the spindle pole, concomitant with (ii) its position relative to the spindle pole in the x-z direction (lateral positioning affects spindle orientation in the plane (x-y) of the monolayer), (iii) the existence of a recently reported spindle tilting correction mechanism during anaphase [41], and (iv) the reintegration of delaminated cells in the monolayer.
Recycling endosomes and therewith associated rab11a have also been shown to contribute to cytokinesis [10]. In addition to mitotic spindle misorientation, a delay in cytokinesis was observed upon the loss of MYO5B expression. Previous studies have demonstrated that rab11apositive recycling endosomes, via the rab11-binding partner FIP3 [42], mediate the delivery of the actin depolymerizing p50RhoGAP [43] and other proteins [44] to the ingressing cleavage furrow during late telophase to promote further ingression and abscission. Also, motor proteins have been implicated in the delivery of recycling endosomes to the cleavage furrow [11,45,46]. Our results demonstrate a role for myosin Vb in cytokinesis, and future studies are needed to unravel the mechanism. Notably, in contrast to the mitotic spindle misorientation, the delay in cytokinesis did not correlate with the presence of giant late endosomes and could not be rescued by the overexpression of rab7, arguing against a role for the giant late endosomes, and possibly in favor of a functional relationship with rab11a, in the cytokinesis phenotype.
We consider it unlikely that the giant late endosome-induced spindle orientation phenotype contributes to the pathogenesis of MVID because the giant late endosomes present in MVID intestine predominantly appear in differentiated villus enterocytes and not in proliferating crypt cells. It should be noted that our results were obtained primarily in cell culture which does not fully recapitulate intestinal physiology. The intestine of the Myo5b knockout mouse may show adaptive physiological responses that involve proliferation of crypt cells, and in vivo conditions and outcomes may therefore be different. The physiological relevance of the mechanisms described here in the in vitro systems thus remains to be elucidated. Future studies should address the possible contribution of oversized endosome-related spindle misorientation to the pathogenesis of human diseases in which aberrantly sized endosomes have been reported [47], which include neurodegenerative diseases [48,49], infectious diseases [21], rare inherited diseases [50][51][52][53], and cancer [54,55], and the potential of modulating rab7 availability as a therapeutic avenue.

Virus production and transduction
Lentiviral particles were created using a second-generation system. Briefly, following the instruction of Fugene (Promega, Madison, WI, USA), the mixture of pLenti-plasmid, vesicular stomatitis virus-glycoprotein, delta-8.91, and Fugene was added in the 6-well plate, and 1 × 10 6 HEK293T cells were seeded on top of the mixture. After overnight incubation, medium was changed with DMEM with 2.5% (v/v) serum and 1% antibiotics. After 48 h, viral particles were harvested and filtered through 0.45-μm filter (GE healthcare, Hatfield, UK). Cells were transduced by filtered virus supernatant with polybrene (8ug/ml) 6 h after seeding. The next day virus medium was changed with selection medium containing appropriate antibiotics.

Generation of cell lines
MYO5B knockout cells were generated using the CRISPR/Cas9 genome editing system. DNA Oligos were the following: sense: caccgGATATCTGGATTCCGTAAGA; antisense: aaacTCTTACGGAATCCAGATATCc.
Virus production and cell transduction procedures were as described above. Transduced cells were selected with puromycin (5 μg/ml, Sigma-Aldrich, St. Louis, MO, USA) 1 day after infection. After 3 days, infected cells were sparsely plated, and individual clones were picked. Finally, the presence of a premature stop codon was checked by sequencing (GATC Company), and myosin Vb expression levels were checked by western blot. Caco2 cells expressing Rab7-WT-GFP or Rab7-GFP-T22N, β-Tubulin-GFP, and Histone2B-mCherry were generated by virus transduction with plenti-plasmids as described above.

Western blotting
Cells were harvested with lysis buffer (100 mM NaCl, 20 mM Tris-HCl [pH 7.6]), Triton X-100 and protease inhibitors (Roche, Almere, the Netherlands). Protein concentration was determined using the BCA protein assay (Sigma-Aldrich, St. Louis, MO, USA). Protein extracts (20 μg) were resolved on an SDS/4-15% polyacrylamide gel, transferred to nitrocellulose membranes, and blocked with Odyssey Blocking Buffer (LI-COR, Lincoln, NE, USA) at room temperature for 1 h. The membrane was incubated with primary antibodies (see S1 Table) at 4˚C for 16 h, washed 3 times with PBS with 0.1% (v/v) Tween, and incubated with the secondary antibody in OBB/PBS buffer at room temperature for 1 h. Following washes with PBS with 0.1% (v/v) Tween, antibody signal was detected using Odyssey Imaging System (LI-COR, Lincoln, NE, USA).

Immunofluorescence and confocal microscopy
For immunofluorescence microscopy, cells were fixed in 4% paraformaldehyde for 20 min and permeabilized with 0.2% (v/v) Triton X-100 at room temperature for 10 min. Cells were washed with PBS at each step. Cells were incubated in PBS with 1% (w/v) BSA at 37˚C for 1 h, followed by overnight incubation with primary antibodies (see S1 Table) in PBS with 1% (w/v) BSA at 4˚C. After washes with PBS, cells were incubated with secondary antibodies in PBS with 1% BSA (w/v) at room temperature for 1 h. Following washes with PBS, cells were mounted in Dako mounting medium and imaged using a fluorescence microscope (Leica, Frankfurt, Germany). In mitotic tilted experiments, cells were cultured in an 8-chambers plate (Lab-Tek II, 155409) and incubated with DMEM medium for 5 days. Cells were stained as above and imaged by TCS SP8 confocal microscope (Leica DMI 6000) (Leica, Frankfurt, Germany) with x-y-z stage.

Electron microscopy
Tissues were fixed in 2% glutaraldehyde in 0.1 M sodiumcacodylate buffer at 4˚C for 24 h. Cells grown in 6-well plates were fixed by adding dropwise an equal volume fixative (2% [v/v] glutaraldehyde, 2% [v/v] paraformaldehyde in 0.1 M sodiumcacodylate buffer). After 10 min, this mixture was replaced by pure fixative at room temperature for 30 min. After postfixation in 1% osmium tetroxide/1.5% potassium ferrocyanide (v/v, 4˚C, and 30 min), tissues and cells were dehydrated using ethanol and embedded in EPON epoxy resin. A total of 60 nm sections were cut and contrasted using 2% (v/v) uranyl acetate in water followed by Reynolds lead citrate. Images were taken with a Zeiss Supra 55 in STEM-mode at 26KV using an external scan generator (Fibics, Canada) yielding mosaics of large area scans at 2.5 nm pixel resolutions.

Immunohistochemistry
Immunohistochemistry was performed on 4 μm-thick tissue sections. The sections were deparaffinized, rehydrated in xylene and graded ethanol solutions. Sodium citrate buffer (10 mM sodium citrate, 0.05% (v/v) Tween-20, [pH 6.0]) was used for antigen retrieval, and slides were incubated overnight with primary antibodies (see S1 Table), followed by incubation with secondary antibodies and detection with horseradish peroxidase and 3,3'-diaminobenzidine. Slides were counterstained with hematoxylin and covered with mounting medium (Dako, Santa Clara, CA, USA).

Treatment cells with different inhibitors
In experiments with vATPase and chloride channel inhibitors, cells were seeded on the coverslips in a 24-well plate and incubated with DMEM medium for 72 h. The vATPase inhibitor Baf A1 (0.1 μM), chloride channel inhibitors NPPB (200 μM), DIDS (500 μM), and furosemide (2 mM; Sigma-Aldrich, St. Louis, MO, USA) were added into the medium. After incubated with chemicals for indicated times, cells were fixed with PFA, followed by mounting of the slides and microscopical examination. In experiments with the PI3K inhibitor Wm (Sigma-Aldrich, St. Louis, MO, USA), cells were cultured as above and incubated with Wm (500 nM) for 1 h before fixation. In mitotic tilting and delamination experiments, cells were cultured in an 8-chambers plate (Lab-Tek II, 155409) (Thermo Scientific, Waltham, MA, USA)and incubated with DMEM medium for 5 days. Cells were treated with Baf A1 (0.1 μM) for 24 h or Wm (500 nM) for 1 h before fixation. Then cells were immunolabeled as described above and imaged with a TCS SP8 confocal microscope (Leica DMI 6000) (Leica, Frakfurt, Germanywith x-y-z stage.

Live cell imaging
Cells expressing β-tubulin-GFP and histone2B-mCherry were seeded in an 8-chambers plate (Lab-Tek II, 155409) (Thermo Scientific, Waltham, MA, USA) and incubated with DMEM medium for 72 h. Then the live cells were observed 24 h with the laser scanning confocal microscope (GE Healthcare Bio-Sciences, Marlborough, MA, USA) with temperature and CO 2 control. Images were collected using a 40× oil objective with GFP-mCherry and DIC channels every 6 to 8 min. Images were processed by Imaris software (Oxford Instruments, Zurich, Switzerland; https://imaris.oxinst.com/downloads).

Quantification and statistical analysis
For live cell imaging and mitotic tilted experiments, images were performed with confocal microscope with x-y-z stage. In other immunofluorescence experiments, images were 2D object with x-y stage. All experiments were performed for at least 3 independent experiments. The measurements of the α-angle in horizontal rotation experiment were obtained with the free ImageJ tools (https://imagej.nih.gov/ij/download.html) in x-y confocal sections. The line of dividing direction was drawn vertically to the chromosome plane through the center in metaphase and drawn through the both the 2 daughter cells' chromosome plane in anatelophase. The β-angle was obtained with the ImageJ tools x-z confocal sections. The indicated vacuole's position was confirmed by bright field images. To estimate the percentage of cells not contacting the substratum, 5 random x-y fields of 4 × 10 4 μm 2 were analyzed in the x-y-z dimensions for 3 independent experiments. For statistical estimation, we used GraphPad Prism version 7.0 software (Graphpad software, San Doego, CA, USA) (https://www. graphpad.com/scientific-software/prism). Graphs represent mean y ± SEM or + SD, as indicated, n represents the numbers of cells analyzed. Statistical significance was determined by unpaired 2-tailed t tests. were surrounded by microtubules (β-tubulin) and stained negative for markers of early/recycling endosomes (EEA1, TfR, rab8a, and rab11) but positive for markers of late endosome/ lysosomes (rab7 and LAMP1). Endosomal markers were often found very close to the vacuoles (arrows), which could be because of their association with the perivacuolar microtubule cytoskeleton. To distinguish whether these markers were truly associated with the vacuoles or located subjacent to the vacuoles, we treated the cells with nocodazole (33 μM) for 2 h. In contrast to the resultant dispersal of early endosomal and recycling endosomal markers, rab7 and LAMP1 remained positive at the vacuoles. Scale bars: 10 μm. EEA1, early endosomal antigen-1; LAMP1, late endosome-associated membrane protein; TfR, transferrin receptor.