Vesicle-mediated transport of ALIX and ESCRT-III to the intercellular bridge during cytokinesis

Cellular abscission is the final step of cytokinesis that leads to the physical separation of the two daughter cells. The scaffold protein ALIX and the ESCRT-I protein TSG101 contribute to recruiting ESCRT-III to the midbody, which orchestrates the final membrane scission of the intercellular bridge. Here, we addressed the transport mechanisms of ALIX and ESCRT-III subunit CHMP4B to the midbody. Structured illumination microscopy revealed gradual accumulation of ALIX at the midbody, resulting in the formation of spiral-like structures extending from the midbody to the abscission site, which strongly co-localized with CHMP4B. Live-cell microscopy uncovered that ALIX appeared together with CHMP4B in vesicular structures, whose motility was microtubule-dependent. Depletion of ALIX led to structural alterations of the midbody and delayed recruitment of CHMP4B, resulting in delayed abscission. Likewise, depletion of the kinesin-1 motor KIF5B reduced the motility of ALIX-positive vesicles and delayed midbody recruitment of ALIX, TSG101 and CHMP4B, accompanied by impeded abscission. We propose that ALIX, TSG101 and CHMP4B are associated with endosomal vesicles transported on microtubules by kinesin-1 to the cytokinetic bridge and midbody, thereby contributing to their function in abscission. Supplementary Information The online version contains supplementary material available at 10.1007/s00018-023-04864-y.

Vesicle transport into the ICB is mediated by molecular motor proteins [3,51]. Microtubule-based transport depends on kinesin and dynein motor proteins, with kinesins mediating plus-end-and dyneins responsible for minus-enddirected transport [52]. Rab11-FIP3 endosomes are transported along microtubules [53] and Rab35 endosomes are moving in the bridge [26], but if this occurs by kinesinmediated transport is not known. Interestingly, bi-directional movement of Rab11 and Rab35 has been reported in early bridges, whereas in late bridges they are mostly stationary [26,[54][55][56]. FIP3 can directly interact with Rab11 and the bi-directional movement of FIP3-positive endosomes is dependent on a switch between plus-end and minus-end motor proteins (KIF5B and dynein) and controlled by Arf6 and JIP4 [54]. In MDCK (Madin Darby Canine Kidney) cells the plus-end-directed motor KIF3A/B mediates the transport of Rab11-FIP5-positive endosomes to the center of the bridge [53]. On the other hand, also actin-dependent motor proteins, such as myosin VI [57], are involved in cytokinesis, and the transport of Rab8-positive vesicles might depend on myosin VI [3]. In human cells the motor proteins for Rab35-positive or secretory vesicle transport into the bridge are not known [3].
Even though functional roles of ALIX and other midbody-associated proteins, such as CHMP4B and TSG101, during cytokinesis have been studied in detail [16,17,58], relatively little is known about the mechanisms and spatiotemporal dynamics by which these proteins are transported into the ICB and eventually to the midbody.
In this study, we address the spatiotemporal dynamics and mechanisms of ALIX and the ESCRT-III subunit CHMP4B transport during cytokinesis. In accordance with Addi et al. [59], we demonstrate a gradual accumulation of ALIX, in co-localization with CHMP4B, at the midbody and toward the abscission site in spiral-like structures. Moreover, our data suggest a highly dynamic recruitment of ALIX and CHMP4B to the midbody and that ALIX and CHMP4Bpositive vesicles undergo directional transport along microtubules to the periphery of and into the ICB, eventually contributing to their midbody recruitment. We further demonstrate that the microtubule motor kinesin-1 promotes transport of ALIX-positive vesicles, the recruitment of ALIX, TSG101 and CHMP4B to the midbody as well as accurate timing of cytokinetic abscission.

Highly dynamic populations of ALIX and CHMP4B gradually accumulate at the midbody and form spiral-like structures
At late stages of cytokinesis, ALIX is well-documented to be recruited to the midbody and to promote CHMP4B midbody recruitment [16][17][18]. During maturation of the cytokinetic bridge the localization of ALIX and CHMP4B changes from the appearance as two parallel stripes at the midbody to a cone-shaped structure [22,59,60]. Furthermore, ALIX and CHMP4B show a high degree of co-localization and eventually both proteins are recruited to the abscission site [59]. We were interested to understand the spatiotemporal dynamics of ALIX recruitment to the midbody in further detail. To do this, we first examined endogenous ALIX at different stages of midbody maturation in HeLa K cells using structured illumination microscopy (SIM) (Fig. 1a). We detected early appearance of endogenous ALIX in dotted structures at the midbody (Fig. 1a, panel I.), followed by the formation of two ALIX-positive ring-like structures adjacent to the midbody (Fig. 1a, panels II.-III.), formation of elongated conelike ALIX-positive structures extending from the midbody (Fig. 1a, panels IV.-V.) and finally appearance of ALIX at the abscission site (Fig. 1a, panel VI. and Fig. 1b). We also monitored a high degree of association between ALIX and CHMP4B during this process by SIM (Fig. 1c). ESCRT-IIIdependent contractile filaments form at the constriction sites in the ICB [8,21,24]. Accordingly, the detailed analysis of our 3D SIM data revealed that the cone-like structures of ALIX and CHMP4B in two-dimensional microscopy correspond to three-dimensional spiral-like structures ( Fig. 1c and Movie 1a). These findings are consistent with the morphological changes of ALIX and CHMP4B at the midbody described previously [22,59]. To investigate the dynamics of ALIX and CHMP4B recruitment to the midbody, we performed FRAP (fluorescence recovery after photobleaching) analysis of ALIX and CHMP4B at the midbody. The FRAP data showed a very fast recovery of both proteins at the midbody, indicating the existence of highly dynamic protein populations of ALIX and CHMP4B ( Fig. 1d and Movie 1b). On the other hand, no substantial recovery occurred at post-mitotic midbody remnants ( Fig. 1d and Movie 1c). Thus, our data suggest that ALIX and CHMP4B gradually accumulate at the midbody in a highly dynamic manner, leading to the formation of spiral-like structures and their eventual accumulation at the abscission site.  1 Gradual accumulation of ALIX and CHMP4B at the midbody and colocalization in spiral-like structures. a Sequential accumulation of ALIX at the midbody. 3D SIM microscopy of fixed cells stained for ALIX (magenta), RacGAP1 (green) and tubulin (grey) at different progressive stages of cytokinesis (I.-VI.). b Recruitment of ALIX to the secondary ingression and abscission sites in fixed cells stained for ALIX (magenta) and tubulin (grey). Inlays show structural changes of ALIX at the midbody, secondary ingression and abscission sites (arrowheads) in (a and b). c Colocalization of ALIX (magenta) and CHMP4B (green) and formation of spiral like structures at the midbody in fixed cells. Representative images show projections of 3D reconstructed SIM data at different visual angles. Cells were stained for ALIX (magenta) and CHMP4B (green). Scale bars in (ac) = 2 µm. In (a-c) > 50 cells from at least three independent experiments were analyzed. See also Movie 1a. d FRAP analysis of ALIX and CHMP4B dynamics at the midbody (red) and in post abscission midbody remnants (grey). Normalized intensities from four independent experiments of cells stably expressing ALIX-mCherry (top) or CHMP4B-GFP (bottom) are plotted (number of FRAP measurements: ALIX-midbody = 12, ALIX-remnant = 11, CHMP4B-midbody = 10, CHMP4B-remnant = 11). See also Movie 1b-c 235 Page 4 of 18 ALIX depletion leads to delayed CHMP4B recruitment and abscission timing as well as altered ICB and midbody morphology ALIX plays an evolutionarily conserved role in promoting cytokinesis [16-19, 61, 62] and depletion of ALIX delays the completion of the abscission process [59,61]. To characterize the cytokinetic effects of ALIX depletion in our cellular system, HeLa K cells, we performed siRNA-mediated depletion of ALIX in these cells ( Fig. 2a and Suppl. Fig. 1a). We analyzed abscission time in live cell imaging experiments in control and ALIX-depleted cells by measuring the time starting from the formation of a stable cytokinetic bridge until the microtubules in the bridge were severed ( Fig. 2b and Suppl. Fig. 1b). Consistent with earlier findings [16,59], abscission was strongly delayed upon ALIX depletion using c Scatter plot showing the time interval between bridge formation and first appearance of CHMP4B-GFP at the midbody (MB) in live cell imaging analysis upon control (ctrl.) and ALIX siRNA (ALIX KD) treatment as indicated (n ≥ 60 cells per treatment from three independent experi-ments; control: 59.9 ± 1.4 min; ALIX KD: 137.8 ± 10.6 min [± SEM]; P < 0.001). d Quantification of the length of the ICB from live cell imaging data (n ≥ 100 cells from three independent experiments; control: 6.33 ± 0.1 µm; ALIX KD: 9.5 ± 0.29 µm [± SEM]; P < 0.001). e Altered morphology of the midbody upon ALIX depletion. Cells were fixed and stained for RacGAP1 (green), CEP55 (magenta) and tubulin (grey). Overview images as well as projections of the midbody region (RacGAP1 and CEP55 staining) are shown. Control cells show a compact midbody. In contrast to control cells, ALIX-depleted cells display enlarged midbodies with filamentous extensions. Scale bars = 5 µm two different siRNAs in HeLa K cells compared to control HeLa K cells ( Fig. 2b and Suppl. Fig. 1b). ALIX directly interacts with and promotes recruitment of CHMP4B to the midbody [17,18]. Consistently, ALIX depletion resulted in substantially delayed CHMP4B midbody recruitment in live cell imaging experiments in HeLa K cells compared to control cells ( Fig. 2c and Suppl. Fig. 1c-d). Accompanied with these phenotypes, ALIX depletion also led to the appearance of elongated cytokinetic bridges (Fig. 2d). The analysis of our high-resolution SIM data also revealed that knockdown of ALIX interfered with the structural properties of the midbody (Fig. 2e). Midbodies stained with an antibody against RacGAP1 (Rac GTPase Activating Protein 1), a component of the centralspindlin complex that forms a core structure at the midbody, exhibited significant morphological alterations in ALIX knockdown cells (Fig. 2e). To eliminate morphological changes caused by variations in the cytokinetic phase, we additionally labeled the cells with antibodies against CEP55, a protein that accumulates at the midbody during advanced stages of cytokinesis as a thin ICB has formed [14,15,63]. In control cells, RacGAP1 appeared as a ring-shaped structure with of a diameter of approximately 1 µm ( Fig. 2e and Movie 2). In ALIX-depleted cells these structures were substantially enlarged and often accompanied with filament-like extensions associated to the midbody structure ( Fig. 2e and Movie 2). Quantification of the midbody morphologies revealed a significant increase in irregular midbody morphologies with filamentous extensions in ALIX-depleted cells (38 ± 4%) compared to control cells (13 ± 3%) (see Fig. 6k). These structural alterations might be a consequence of the delayed abscission process associated with an erroneous regulation of mechanical forces within the cytokinetic bridge. Altogether, ALIX depletion in HeLa K cells resulted in significantly delayed CHMP4B midbody recruitment and cytokinetic abscission, increase in ICB length as well as alterations in the midbody morphology.

ALIX and CHMP4B localize to vesicular structures transported along microtubules
In order to elucidate how ALIX is transported to different cellular destinations, particularly to the cytokinetic bridge and the midbody, we performed live cell imaging of HeLa K cells stably expressing fluorescently labelled ALIX and/ or CHMP4B. In earlier studies, ALIX and CHMP4B have been detected in dotted, vesicle-like structures in interphase cells [64]. Here, we find that such ALIX-positive vesicular structures are dynamic and transported along the microtubule network in interphase cells (Fig. 3a, upper panel and Movie 3a). Accordingly, microtubule depolymerization by nocodazole treatment led to aggregation and immobilization of ALIX-positive structures in interphase cells (Fig. 3a, lower panel and Movie 3a). We found that the ALIX-positive vesicles were often also positive for CHMP4B in interphase cells, and consistently, we could detect CHMP4B co-transport with ALIX (Fig. 3b, upper panel). After it became evident that ALIX and CHMP4B localized to vesicle-like structures that are transported along microtubules, we investigated their transport in cells undergoing cytokinesis. As in interphase cells, ALIX and CHMP4B also localized to vesicular structures in cytokinetic cells and both proteins were detected to be co-transported over long distances to the periphery of the cytokinetic bridge (Fig. 3b, lower panel and Movie 3b). The existence of endosomal vesicles in the cytokinetic bridge has been previously documented and their crucial role in protein and membrane transport during cytokinesis is well established [36,37,39,50,65]. Consistently, we also detected vesicles in the lumen of the cytokinetic bridge by scanning transmission electron microscopy (STEM) tomography ( Fig. 3c and Movie 3c). To test whether the intracellular accumulations of ALIX and CHMP4B were associated to vesicular structures we treated HeLa K cells stably expressing fluorescently labelled ALIX and CHMP4B cells with CellBrite ® Steady 650 membrane dye. In live cell imaging, we observed co-localization and co-transport of ALIX and CHMP4B together with CellBritepositive vesicles in interphase cells (Movie 3d) as well as to the periphery of the ICB in cytokinetic cells (Movie 3e). Co-localization of ALIX and CHMP4B with CellBrite in the ICB was also confirmed by super-resolution imaging (Fig. 3d). Thus, in interphase and cytokinetic cells, ALIX co-localizes with CHMP4B on vesicular structures and both proteins are transported on such structures in a directed and microtubule-dependent manner.

ALIX-and CHMP4B-positive vesicles can be transported into the ICB and to the midbody and ALIX is partially co-transported with Rab11
Importantly, in live-cell imaging experiments, ALIX-and CHMP4B-positive vesicles were detected to be transported to the periphery of the ICB as described above (Movie 3e). We also detected partial transport of ALIX-or ALIX-and CHMP4B-positive vesicles into the cytokinetic bridge and visualized recruitment of such vesicles to the midbody in both live cell imaging and super-resolution microscopy of fixed cells (Fig. 4a-c and Movies 4a-d). Interestingly, we even observed transport of ALIX-positive vesicles into postabscission bridges, from the side where the second abscission site has not yet undergone abscission (Movie 4b). We further performed Proximity Ligation Assays (PLAs), which revealed close proximity between endogenous ALIX and CHMP4B, both in the cell bodies of cytokinetic cells as well as in the ICB and at the midbody (Fig. 4d, left panel and Suppl. Fig. 3a, b), whereas such signals were clearly reduced in ALIX-depleted cells (Fig. 4d, right panel) or in cells treated with only one primary antibody (Suppl. Fig. 3a, b). Super-resolution microscopy of fixed cells furthermore confirmed proximity of endogenous ALIX and CHMP4B on vesicles both in the cell body as well as in the cytokinetic bridge and at the midbody (Suppl. Fig. 1e and Movie 4e). Interestingly, sometimes ALIX seemed to move in a bi-directional manner in and out of the ICB or in the periphery of the bridge (Movie 4f). Such movement has also been described for Rab11-FIP3 endosomes during early stages of the ICB of cytokinetic cells [25,26,37,54,55]. Interestingly, in live cell imaging experiments we found co-transport of ALIX and Rab11, first to the periphery (Movie 4f) and finally into the ICB ( Fig. 4e and Movie 4h). We also detected presence of endogenous ALIX in close association with Rab11 both in the cell body and in the cytokinetic bridge using both PLA ( Fig. 4f and Suppl. Fig. 3) and SIM in fixed cells (Suppl. Fig. 2a and Movie 4i). Furthermore, FRAP analysis of Rab11 at the midbody showed dynamics resembling those of ALIX and CHMP4B (Suppl. Fig. 2b). Besides Rab11-positive vesicles, Rab35 endosomes also promote cytokinesis as described above [25,26,38,47,49]. Our FRAP analysis showed different dynamics of Rab35 at the midbody than of Rab11 (Suppl. Fig. 2b). Similarly, we could neither detect a high degree of co-transport between ALIX and Rab35 in interphase cells (Movie 4j) nor in the membrane dye (blue) and stained for ALIX (magenta) and CHMP4B (green). The projection of an ICB, highlighting the midbody region, is depicted in the image above, while the central image provides the view of a single section. The separate channels of the highlighted region are displayed below. ALIX and CHMP4B signals are associated to membrane-containing vesicles, as indicated by arrowheads (see also Movies 3d and 3e). Scale bar = 3 µm ICB of dividing cells (Movie 4k). Our data therefore support the assumption that ALIX transport seems to depend more on Rab11 than on Rab35. Thus, the above data suggest that ALIX and CHMP4B are closely associated on vesicles and that directional transport of these vesicles occurs, at least partially, via a Rab11-FIP3-mediated process to the ICB.

TSG101 associates with vesicles and is transported together with ALIX into the ICB
During cytokinesis the ESCRT-III component CHMP4B is recruited to the midbody either directly via ALIX or via an ESCRT-I/TSG101-ESCRT-II-CHMP6-dependent mechanism [16][17][18][19]. Thus, we investigated how the ESCRT-I subunit TSG101 is transported to the midbody. Surprisingly, live cell imaging also revealed co-localization and co-transport between TSG101 and ALIX, both in non-dividing cells (Movie 4l) and in the ICB of dividing cells ( Fig. 4g and Movie 4m). Consistently, PLA showed proximity of endogenous of ALIX and TSG101 both in the cell body and in the ICB of cytokinetic cells ( Fig. 4h and Suppl. Fig. 3). In addition, knockdown of ALIX resulted in significantly delayed recruitment of TSG101 to the midbody as compared to its recruitment in control cells ( Fig. 4i and Movie 4n). This indicates that TSG101 is associated to vesicular structures and that at least a fraction of TSG101 is transported together with ALIX into the ICB, with this recruitment partially being dependent on ALIX.
ALIX is co-transported with the kinesin-1 motor protein KIF5B into the ICB Finally, we examined by which mechanism ALIX is transported to the midbody. Directed intracellular protein and vesicle transport is mediated via specialized motor proteins that move along the cellular actin cytoskeleton or the microtubule network [66,67]. As our data indicated a microtubule-directed transport of ALIX towards and into the cytokinetic bridge, we investigated the role of kinesins [68] in this process. This superfamily of microtubule-associated motor proteins includes 14 families of kinesins that mediate an ATP-dependent transport of different cargo along microtubules, and several kinesins have been identified to play a role in mitosis and cytokinesis [52]. In particular, we focused on kinesin-1, as this kinesin is a major motor for anterograde transport towards the plus-end of microtubules [69]. Furthermore, kinesin-1 is the motor protein for the transport of Rab11-FIP3 vesicles in the cytokinetic bridge [54]. The native conventional kinesin-1 holoenzyme exists as a tetramer consisting of two kinesin heavy chains (KHCs) and two kinesin light chains (KLCs) [70]. In mammals, three genes (KIF5A, KIF5B, and KIF5C) encode KHC (kinesin-1) isoforms. KIF5A and KIF5C are neuron specific, whereas KIF5B is ubiquitously expressed [71,72]. Thus, we focused on KIF5B and investigated its role in the transport of ALIX-positive vesicles to the midbody. Live cell imaging revealed a strong co-localization and transport of KIF5B with ALIX-positive vesicles to the periphery of the ICB (Fig. 5a and Movie 5a). High-resolution microscopy demonstrated proximity of endogenous ALIX and KIF5B ( Fig. 5b and Movie 5b) as well as of ALIX and the kinesin light chain KLC1 (Fig. 5c and Movie 5c) in vesicle-like structures in the cytokinetic bridge. Additionally, PLAs confirmed close ALIX-KIF5B ( Fig. 5d and Suppl. Fig. 3a, b) and ALIX-KLC1 ( Fig. 5e and Suppl. Fig. 3a, b) association in the ICB and/or at the midbody. Altogether, these data suggest a proximity of ALIX with kinesin-1 motor protein KIF5B and light chain KLC1 and thus kinesin-1-mediated ALIX transport on vesicles during late stages of cytokinesis.

KIF5B promotes ALIX, CHMP4B and TSG101 midbody recruitment and facilitates accurate abscission timing
KIF5B has been shown to promote vesicle transport to the midbody and completion of cytokinesis [54,73,74]. Based on the presence of ALIX and KIF5B co-transport to the ICB, we asked whether KIF5B plays a role in the transport of ALIX to the midbody and in the completion of cytokinesis. We therefore analyzed the effect of KIF5B knockdown ( Fig. 6a and Suppl. Fig. 4a) on the abscission time and the recruitment of ALIX to the midbody. Knockdown of KIF5B using two different siRNAs in HeLa K cells led to a strong delay in the abscission process ( Fig. 6b and Suppl. Fig. 4b-c). This is in accordance with previous reports demonstrating that depletion of KIF5B strongly delays abscission in chondrocytes [74]. Interestingly, in FRAP as well as in live cell imaging experiments, the dynamics at and recruitment of ALIX to the midbody were significantly delayed upon KIF5B depletion as compared to control cells, respectively (Fig. 6c, d and Movie 6a). Importantly, the ALIX protein levels were similar in control HeLa K cells and after KIF5B depletion, as detected by Western blot analysis (Fig. 6a). Moreover, CEP55 protein levels were not reduced in KIF5B-depleted cells (Suppl. Fig. 4d), and KIF5B protein levels were similar in both control and ALIX-depleted HeLa K cells (Suppl. Fig. 4e).
Given the co-transport detected between ALIX and CHMP4B or TSG101 above (Fig. 4b, f and Movies 4b, 4c and 4m), we also investigated the effect of KIF5B knockdown on the midbody recruitment of CHMP4B or TSG101. Importantly, KIF5B depletion resulted in significantly delayed recruitment of both CHMP4B (Fig. 6e and Movie 6b) and TSG101 (Fig. 6f and Movie 6c) to the midbody. In line with the delayed midbody recruitment of ALIX, we observed an effect of KIF5B depletion on the motility of ALIX-positive vesicles in cytokinetic cells (Fig. 6g-i). Compared to control cells, KIF5B depletion significantly impaired the total distance travelled by ALIX-positive vesicles (Fig. 6g) as well as the mean and maximum speed of ALIX-positive vesicles (Fig. 6h, i). In addition, similar to the effects induced by ALIX knockdown, KIF5B depletion also resulted in structural alterations of the midbody as revealed by 3D SIM (Fig. 6j). Visualization of RacGAP1 of the centralspindlin complex in cells positive for CEP55 showed a significant increase in enlarged midbodies with irregular shapes and elongated filamentous structures in KIF5B-depleted cells as compared to midbodies in control cells (Fig. 6j, k and Movie 6d). Consistent with the observed phenotypes upon KIF5B depletion, knockdown of the kinesin-1 adaptor KLC1 using two different siRNAs (Suppl. Fig. 4f) also resulted in a significant abscission delay compared to control cells (Suppl. Fig. 4g). Similarly, KLC1 depletion resulted in a significant increase of irregularly shaped midbodies compared to control cells (Fig. 6k, Suppl.  Fig. 4h and Movie 6e). Both ALIX-and KIF5B-depleted cells displayed a significantly increased appearance of multinucleate cells, in line with the strong abscission delay detected in these cells (Suppl. Fig. 4i). In summary, these data show an important role of KIF5B in mediating cytokinetic abscission by enabling directed transport of proteins required for abscission into the ICB and to the midbody. In particular, our data provide evidence that the kinesin-1 motor protein KIF5B promotes transport and recruitment of ALIX, CHMP4B and TSG101 to the midbody.

Discussion
Cytokinesis is a cellular process that demands dramatic morphological reorganizations and is associated with the precise temporal and spatial recruitment of a multitude of different proteins, including ALIX and its associated proteins [11,19,60,75]. ALIX is involved in a variety of cellular processes at different cellular localizations [76][77][78]. This indicates a highly organized and directed transport of ALIX to conduct a regulated recruitment to the sites of action. However, by which mechanisms cellular transport of ALIX, and in particular, how spatiotemporal recruitment of ALIX to the midbody during cytokinesis occurs, has remained unclear. Here, we present data that, to our knowledge, show a previously uncharacterized directed and kinesin-1-dependent transport of ALIX-positive vesicles along microtubules to the periphery of and into the ICB, contributing to the accumulation of ALIX at the midbody.
In line with the crucial functional role in cytokinetic abscission, we ( Fig. 2b and Suppl. Fig. 1b-d), and others [17,19,59] observed a strong delay in abscission upon ALIX depletion and the presence of mitotic defects, such as multinucleation (Suppl. Fig. 4i). This can be explained by the fact that ALIX recruits further downstream proteins, such as CHMP4B [17][18][19]. Accordingly, using live cell imaging, we detected strongly delayed recruitment of CHMP4B to the midbody upon ALIX depletion (Fig. 2c). Furthermore, the delayed abscission process in ALIX-depleted cells was accompanied by elongated ICBs (Fig. 2d). Elongated ICBs upon delayed abscission might be a general phenomenon as it has also been documented as a result of functional inhibition of other cytokinesis-regulating proteins [79]. Interestingly, high-resolution microscopy and 3D reconstruction also revealed interference with the structural integrity of the midbody upon ALIX depletion (Figs. 2e and 6k). Our findings are in line with Carlton et al. who also reported an accumulation of aberrant midbodies after ALIX knockdown [18]. Two scenarios seem plausible to us to explain this phenotype. First, ALIX might be needed as a scaffolding and stabilizing component at the midbody. Indeed, a recent publication has shown that ALIX exists at the midbody in complex with syndecan-4, syntenin and CHMP4B [59]. This complex couples the ESCRT-III machinery to the plasma membrane and therefore stabilizes ESCRT-III at the abscission site and ensures accurate abscission timing [59]. Alternatively, the absence of ALIX and the accompanied delay of cytokinesis and physical prolongation of ICBs might lead to a deregulation of the mechanical forces affecting the midbody. Certainly, cytokinesis requires precise spatiotemporal regulation of mechanical forces [22,[79][80][81][82][83]. Significantly, similar to the ALIX-deficient phenotype, depletion of KIF5B or KLC1 also led to abnormal alterations of the midbody (Fig. 6j, k and Suppl. Fig. 4h). KIF5B functions as a motor protein and KIF5B accumulates in the ICB adjacent to the midbody at late stages of cytokinesis of chondrocytes [74]. In summary, the precise spatiotemporal delivery of certain midbody-associated proteins seems to be essential to ensure balanced mechanical forces in the bridge and to maintain the structural integrity of the midbody. Fig. 4 Directed transport of ALIX and CHMP4B to the midbody. a ALIX transport into the ICB and recruitment to the midbody. Selected frames at indicated time points from time-lapse imaging of cytokinetic cells stably expressing ALIX-mCherry (magenta) upon addition of SiR-tubulin (green). Arrowheads indicate transport of individual vesicles at given time points. Scale bar = 5 µm. b Detection of endogenous ALIX in the cytokinetic bridge. SIM micrograph of fixed cells stained for ALIX (magenta), RacGAP1 (green) and tubulin (blue) showing ALIX in vesicular structures in the ICB, at the abscission site and at the midbody (arrowheads from right to left). Scale bar = 1 µm. c ALIX co-transport with CHMP4B into the ICB. Selected frames at indicated time points from time-lapse imaging of cytokinetic cells stably expressing CHMP4B-GFP (green) and ALIX-mCherry (magenta) upon addition of SiR-tubulin (blue). Arrowheads indicate transport of individual vesicles at given time points. Scale bar = 2 µm. d Detection of endogenous ALIX/CHMP4B proximity by fluorescent microscopy using Duolink ® PLA. Fluorescent dots (red) represent ALIX in close proximity with CHMP4B. In control cells dots can be detected in the cell body as well as in the ICB, stained by tubulin (grey), and at the midbody. ALIX knockdown in cells leads to a substantial decrease in the number of dots. Scale bars = 10 µm. e Co-transport of ALIX and Rab11. Selected frames from a time-lapse microscopy of cytokinetic cells expressing Rab11-GFP (green) and ALIX-mCherry (magenta) with SiRtubulin (blue). Arrowheads of same color indicate transport of individual vesicles at given time points. Scale bar = 3 µm. f Detection of endogenous ALIX/Rab11 proximity by fluorescent microscopy using Duolink ® PLA. Fluorescent dots (red), which represent ALIX in close proximity with Rab11, can be detected in the cell body as well as in the ICB, stained by tubulin (grey), and at the midbody. Scale bars = 10 µm. g Co-transport of ALIX and TSG101. Selected frames from a time-lapse microscopy movie of cytokinetic cells expressing GFP-TSG101 (green) and ALIX-mCherry (magenta) with SiRtubulin (blue). Arrowheads indicate transport of an individual vesicle at given time points. Scale bar = 3 µm. h Detection of endogenous ALIX/TSG101 proximity by fluorescent microscopy using Duolink ® PLA. Fluorescent dots (red), which represent ALIX in close proximity with TSG101, can be detected in the cell body as well as in the ICB, stained by tubulin (grey). Scale bars = 10 µm. ALIX is an ESCRT-III-associated protein involved in a variety of ESCRT-III-dependent cellular processes, including cytokinesis. In these processes, ALIX participates in recruiting CHMP4B to the sites of action [17,18,78,[84][85][86]. During cytokinesis ALIX promotes recruitment of CHMP4B to the midbody (discussed above) and subsequently CHMP4B polymerizes into helical filaments that form spiral-like structures toward the site of abscission [8,21,23,59]. We detected a high degree of ALIX/CHMP4B co-localization at such spirals at late stages of cytokinetic abscission, identifying that these spirals are also positive for ALIX (Fig. 1c), which is in line with findings by Addi et al. [59]. Generally, it is assumed that the recruitment of CHMP4B to the midbody occurs in dependency of ALIX, but only after its appearance. However, our data show proximity or association of ALIX and CHMP4B already outside of the cytokinetic bridge (Figs. 3b and 4c, d), and that subsequently both proteins can be co-transported along microtubules in vesicles to the periphery of and partially into the cytokinetic bridge and then finally accumulate at the midbody (Fig. 4c, d, Suppl. Fig. 1e and Movies 3e and 4c). Accordingly, in our experiments we detected an almost identical temporal recruitment of ALIX and CHMP4B to the midbody (Fig. 6d, e). Importantly, at late stages of cytokinesis, ALIX and CHMP4B showed a continuous high degree of co-localization at progressive stages of midbody maturation and spiral formation as detected by super-resolution microscopy ( Fig. 1c and Movie 1a). Thus, these data propose a new temporal and mechanistic progression in the recruitment of both ALIX and CHMP4B to the midbody.
Live-cell imaging and super-resolution microscopy showed ALIX localized to endosomal vesicles and co-transported together with CHMP4B in interphase cells as well as in cytokinetic cells. In particular, we were able to visualize transport of ALIX/CHMP4B-positive vesicles first to the periphery and subsequently partially into the cytokinetic bridge and to the midbody (Figs. 3b, 4c, d and Movies 3e and 4c, d). Vesicular transport of ALIX and CHMP4B into the ICB enables precision in the regulation of their spatial and temporal targeting towards the midbody. An alternative to directional and motor protein-mediated transport is the diffusion of freely accessible ALIX molecules to the  ALIX and KLC1 (c), respectively. Scale bars = 1 µm. d Detection of endogenous ALIX/KIF5B proximity by fluorescent microscopy using Duolink ® PLA. Fluorescent dots (red), which represent ALIX in close proximity to KIF5B, can be detected in the cell body as well as in the ICB, stained by tubulin (grey). e Detection of endogenous ALIX/KLC1 proximity by fluorescent microscopy using Duolink ® PLA. Fluorescent dots (red), which represent ALIX in close proximity to KLC1, can be detected in the cell body as well as in the ICB, stained by tubulin (grey), and at the midbody. Scale bars = 10 µm sites of action. Indeed, we cannot exclude the possibility that ALIX diffusion might occur in addition to vesicle-associated transport, especially in the cytokinetic bridge and to the midbody. However, a mainly diffusion-driven ALIX distribution would lack complex regulatory mechanism and fast diffusion of vesicle-associated ALIX can physically be excluded. Furthermore, in interphase cells, microtubule dissociation led to aggregation and inhibition of ALIX transport (Fig. 3a), demonstrating the dependency of an intact microtubule network for ALIX transport.
We identified kinesin-1 as an important motor protein for vesicle-associated ALIX, CHMP4B and TSG101 transport to the midbody in HeLa K cells (Figs. 5 and 6). KIF5B is the major kinesin-1 motor in non-neuronal mammalian cells and kinesin-1 possesses a crucial role in protein and membrane transport during cytokinesis, in particular the midbodydirected transport of Rab11-FIP3-positive vesicles [39,50,54]. Interestingly, the proximity and co-transport of ALIX and Rab11 that we found particularly in the periphery of the cytokinetic bridge, but also in the bridge, suggests that a certain fraction of ALIX is co-localized and co-transported on Rab11-FIP3-positive vesicles (Fig. 4e, f, Suppl. Fig. 2a and Movie 4h-i). On the other hand, it seems likely that ALIX is transported also independently of Rab11, possibly by recruitment to vesicles that lack Rab11. Depletion of KIF5B did not entirely inhibit the recruitment of ALIX to the midbody, but significantly delayed its appearance at the midbody (Fig. 6c, d) and interfered with the dynamics of ALIX-positive vesicles (Fig. 6g-i). However, as we did not obtain a complete depletion of KIF5B, small amounts of KIF5B might be sufficient to mediate a certain ALIX transport. The significantly delayed recruitment of ALIX to the midbody upon KIF5B depletion however suggests that kinesin-1 is an important motor protein for ALIX transport to the midbody during cytokinesis.
In addition to the proximity of ALIX with the kinesin heavy chain KIF5B (Fig. 5b, d), we also observed proximity of ALIX with the kinesin light chain (KLC) KLC1 (Fig. 5c, e) in the intercellular bridge and at the midbody. The kinesin-1 family can consist of four different light chains, which mediate substrate binding [87]. Additional studies need to be conducted to investigate the extent to which ALIX might associate with the other KLCs and to answer how ALIX or ALIX-positive vesicles are tethered to kinesin-1. To our knowledge, no direct interaction of ALIX with KLCs has yet been documented. Thus, to fully understand mechanisms and regulation underlying the recruitment of ALIX to the midbody, it is important to decipher the detailed interactions between ALIX, kinesin-1, the specific KLCs and further proteins involved.
Our data showed that KIF5B promotes the recruitment of CHMP4B to the midbody (Fig. 6e), which is consistent with its co-transport with ALIX and a similarly delayed recruitment of ALIX (Fig. 6d). Interestingly, KIF5B also promoted accurate recruitment timing of TSG101 to the midbody (Fig. 6f) and ALIX was not only co-localized and co-transported with CHMP4B, but also with TSG101 (Fig. 4g, h), in the ICB. Thus, ALIX, CHMP4B and TSG101 are all, at least partially, transported by a KIF5B-dependent mechanism to the midbody.
It seems very likely that beside a KIF5B-mediated ALIX and CHMP4B transport, other mechanisms may occur in parallel or in compensation to reduced KIF5B levels. At least a certain fraction of ALIX and CHMP4B proteins are associated to intracellular vesicles that are transported along microtubules, and this kind of transport depends on motor proteins. Even after KIF5B depletion ALIX can be found associated to motile vesicles (Fig. 6g-i), indicating that either very small amounts of KIF5B are sufficient to maintain ALIX vesicle transport or that their transport can be mediated by other motor proteins. In addition to kinesin-1, other kinesin families have been shown to mediate cargo transport in the ICB and to the midbody, such as kinesin-2 and kinesin-3 [33,53,88]. We assume that besides the recruitment of ALIX and CHMP4B to vesicles, large cytoplasmic amounts of these proteins exist. Thus, in addition to long distance transport along microtubules, local recruitment could also be mediated by cytosolic diffusion. This type of short distance transport could enable a very fast protein recruitment, as we observe during cytokinesis (Fig. 1d). Consequently, besides vesicle-mediated transport and recruitment of ALIX and CHMP4B to the midbody, direct recruitment by diffusion would also be feasible. In this case, it would be possible to ensure a high local cytoplasmic concentration of ALIX and CHMP4B by targeted transport of ALIX/CHMP4B-positive vesicles to the periphery of and/ or into the ICB. Indeed, during cytokinesis a large number of ALIX and CHMP4B-positive vesicles were found in the direct periphery of the ICB (Figs. 3d, 4a, Suppl. Fig. 1e and Movies 3e and 4a-d). To understand which other mechanisms contribute to the recruitment of ALIX and CHMP4B to the midbody, further experiments are required.
Generally, it is assumed that the midbody is formed by sequential recruitment of the different associated proteins. In contrast to this hypothesis, our findings indicate that ALIX, TSG101 and CHMP4B can be transported in close proximity on identical endosomes on the one hand (Fig. 4c,  d, g, h, Movies 4c-d and 4m), and on the other hand, that knockdown of ALIX delays the recruitment of CHMP4B and TSG101 to the midbody (Figs. 2c and 4i). Thus, it seems possible that recruitment of these proteins might not occur sequentially and separately to the midbody, but that they might already form complexes outside the bridge, which are then transported as a unit to the midbody.
Taken together, our data uncover kinesin-1-mediated directed transport of ALIX, TSG101 and CHMP4B-positive vesicles to and partially into the cytokinetic bridge, which is necessary to ensure normal execution of cytokinesis. Accordingly, depletion of ALIX or KIF5B interferes with the structural integrity of the midbody, the recruitment of midbody-associated proteins, including TSG101 and CHMP4B, and with the abscission process. Further studies are needed to elucidate whether these mechanisms apply in other cell types and species and in ALIX, CHMP4B and TSG101 transport to other cellular targets.

Cell culture, plasmids and transfection
Human HeLa ''Kyoto'' (HeLa K) cells were maintained in DMEM (GIBCO) supplemented with 10% FBS, 100U/ ml penicillin, and 100 mg/ml streptomycin at 37 °C under 5% CO2. Stable cell lines expressing fluorescently labelled ALIX, CHMP4B or TSG101 were previously described [16,89]. For transient transfection cells were incubated with mixture of FuGene 6 (Promega) and the plasmid of interest, using a ratio of 3:1 of Fugene 6 to DNA and incubated for 24-48 h. In most cases, cells were transfected in MatTek 3.5 cm dishes using 6 µl of Fugene 6 and 2 µg of DNA. Before cells were used for further analysis the medium was exchanged. The pEGFP-ALIX [59], pEGFP-Rab35 and pmCherry-Rab35 [90,91] plasmids were a kind gift from Dr. Arnaud Echard, the plasmid expressing Rab11-RFP was kindly provided by Dr. Kay Schink and the plasmid expressing mCitrine-KIF5B was kindly provided by Dr. Eva M. Wenzel. The plasmid expressing ALIX-mCherry [89], has been described previously.

Live-cell microscopy
Live-cell imaging was performed on a DeltaVision OMX V4 microscope equipped with three PCO.edge sCMOS cameras, a solid-state light source and a laser-based autofocus. For long-term live-cell microscopy (12-16 h; analysis of cell abscission, protein recruitment during cytokinesis and length of ICBs) a DeltaVision microscope (Applied Precision) equipped with an Elite TruLight Illumination System, a CoolSNAP HQ2 camera and a 60 × Plan Apochromat (1.42 numerical aperture) lens was used. For temperature control during live observation, the microscope stage was kept at 37 °C by a temperature-controlled incubation chamber. Cells were imaged in live-cell imaging solution (Invitrogen #A14291DJ) supplemented with 20 mM glucose, 100 U/ml penicillin, and 100 mg/ml streptomycin at 37 °C. Time-lapse images (5-10 z-sections, 0.5-2 µm separation) were deconvolved using SoftWoRx software (Applied Precision, GE Healthcare) and processed with FIJI/ImageJ. For visualization of microtubules, cells were pre-treated for 1-3 h with 75 nM SiRtubulin (Spirochrome). For membrane staining, cells were pre-treated 16-20 h with CellBrite ® 650 according to the manufacturers' instructions (Biotium). For the analysis of ALIX-vesicle motility, cytokinetic cells were visualized on a OMX 4 V microscope (7 min total; 5 s/frame), processed as described above and analyzed with the TrackMate plugin in ImageJ [92].

Structured illumination microscopy (SIM)
For 3D-SIM (structured illumination microscopy) cells were seeded on coverslips and fixed in 4% EM-grade paraformaldehyde for 15 min [93] and permeabilized with 0.1% Triton X-100 in PBS for 5 min. For protein staining, specific primary antibodies were used as defined in the corresponding Figure legends. Coverslips were mounted in ProLong™ Gold or ProLong™ Glass (ThermoFisher). 3D-SIM imaging was performed on a DeltaVision OMX V4 system (Applied Precision) equipped with an Olympus 60 × numerical aperture (NA) 1.42 objective, three PCO. edge sCMOS cameras and 405, 488, 568 and 642 nm diode lasers. Z-stacks covering the whole cell were recorded with a Z-spacing of 125 nm. A total of 15 raw images (five phases, three rotations) per plane were collected and reconstructed by using SoftWoRx software (Applied Precision), processed in ImageJ/Fiji [94] and three-dimensional reconstruction was calculated and visualized by icy imaging software [95].

Fluorescence recovery after photobleaching (FRAP)
For FRAP experiments we used a DeltaVision OMX microscope (Applied Precision, GE Healthcare) with a PlanApo 60 × /1.4 NA oil objective. The cells were imaged at a frame rate of 1 frame per second over a period of 40 s (3 s pre-bleach). Spots of the radius a = 1.7 μm were bleached with a 488-nm laser at 50% laser intensity for 1 s. Intensity change over time was collected with a 488-nm laser for GFP-conjugated constructs and a 561-nm laser for mCherry or RFP-constructs, respectively. FRAP data were analyzed and recovery curves were plotted after background deduction, normalization of fluorescence intensities (the fluorescence intensities at the last frame before bleaching was set to 100%).

STEM tomography
Cells were grown on coverslips and then fixed with 2% glutaraldehyde in 0.1 M PHEM buffer (60 mM PIPES, 25 mM HEPES, 2 mM MgCl2, 10 mM EGTA, pH 6.9) for 1 h. Postfixation was done in 1% OsO 4 and 1.5% KFeCN in the same buffer. Samples were further stained en bloc with 4% aqueous uranyl acetate for 1 h, dehydrated in graded ethanol series and embedded with Epon-filled gelatine capsules (EMS Polysciences Inc.) placed on top of the coverslip. After polymerization serial sections (750 nm) were cut on an Ultracut UCT ultramicrotome (Leica, Germany) and collected on formvar-coated slot grids. Samples were observed in a Thermo Scientific ™ Talos ™ F200C microscope at 200 kV using the bright field detector for STEM (scanning transmission electron microscopy) imaging with microprobe mode. The convergence angle of the scanning beam was 1.7 mrad, condenser aperture of 70 nm and a camera length of 530 mm. For STEM tomography image series were taken at − 56° to 56° tilt angles with 2° increment and a pixel size of 2.25 nm. Tomograms were computed using weighted back projection using the IMOD package. Display of tomogram slices was also performed using IMOD software version 4.9.3.

Statistical analysis
Statistical analysis was carried out in Graphpad Prism (Graphpad Software). Student's t-test was used to compare two groups. ANOVA was used to compare multiple groups and Holm-Sidak was used to correct for multiple comparisons. The threshold for significance was set at P = 0.05. All comparisons made are reported-regardless of significance. Comparisons in the Figures are indicated as n.s.: P > 0.05, *: P < 0.05, **: P < 0.01, ***: P < 0.001. Data availability All data generated or analyzed during this study are included in this article (and its Supplementary Information files) and are available from the corresponding author on reasonable request.

Conflict of interest
The authors declare no competing financial interests.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http:// creat iveco mmons. org/ licen ses/ by/4. 0/.