Engulfment of the midbody remnant after cytokinesis in mammalian cells

Post-mitotic cells do not simply break apart. Decades of observations have revealed that cytokinesis proceeds through complex and stereotyped cell shape changes (Barr and Gruneberg, 2007; Deschamps et al., 2013; Eggert et al., 2006; Fededa and Gerlich, 2012; Green et al., 2012; White and Glotzer, 2012), and systematic RNA interference (RNAi)-based screens have revealed that >100 different proteins are directly or indirectly involved in this process in animal cells (Echard et al., 2004; Eggert et al., 2004; Skop et al., 2004). Cytokinesis begins with the contraction of an actin–myosinII-based furrow and terminates with the cut of the intercellular bridge connecting the two sister cells, through a process called abscission. The intercellular bridge is composed of a dense array of microtubules, the plus ends of which overlap in a central region called the midbody (or Flemming body), first described in the late 1800s. The midbody is particularly rich in proteins, and it appears as a dense and characteristic structure both in phase contrast microscopy and in electron microscopy (Byers and Abramson, 1968; Mullins and Biesele, 1973; Mullins and McIntosh, 1982). A number of proteins, such as anillin (Hu et al., 2012; Kechad et al., 2012; Lekomtsev et al., 2012), strongly connect the membrane to the intercellular bridge and form a ring-like structure surrounding the midbody.

The intercellular bridge is a strikingly active and dynamic structure, and several events must be coordinated before abscission can occur. First, the protein and lipid composition of the plasma membrane is remodeled, which involves vesicular delivery through several parallel trafficking pathways [e.g. Rab35, Rab11-FIP3 (also known as Rab11 family interacting protein 3), phosphatidylinositol 3-phosphate (PI3P), exocyst] (Echard, 2012b; Fielding et al., 2005; Goss and Toomre, 2008; Gromley et al., 2005; Kouranti et al., 2006; Montagnac et al., 2008; Neto et al., 2011; Sagona et al., 2010; Wilson et al., 2005). For instance, the Rab11- and the Rab35-regulated pathways prevent F-actin accumulation at the plasma membrane, and are both required for normal abscission (Chesneau et al., 2012; Dambournet et al., 2011; Echard, 2012a; Schiel et al., 2012). Then, microtubules must be locally severed on the midbody side, which requires the enzymatic activity of spastin (Connell et al., 2009; Guizetti et al., 2011; Lafaurie-Janvore et al., 2013). Finally, ‘endosomal sorting complex required for transport’ (ESCRT)-dependent helical filaments extend from the midbody, creating a constriction zone where abscission will take place (Bastos and Barr, 2010; Capalbo et al., 2012; Carlton et al., 2012; Carlton and Martin-Serrano, 2007; Elia et al., 2011; Guizetti et al., 2011; Lafaurie-Janvore et al., 2013; Morita et al., 2007). The involvement of the ESCRT complexes was a major finding, as ESCRT is believed to be the final machinery that drives the fusion of the plasma membrane in a zone that is free of actin and microtubules (Caballe and Martin-Serrano, 2011; Fededa and Gerlich, 2012). Taken together, the midbody and its adjacent regions are crucial, and they serve as a platform to recruit the proteins that are necessary for abscission.

The fact that the bridge is cut not in the middle of the midbody but, instead, adjacent to it, has interesting consequences. Indeed, after the bridge is cut, a midbody remnant (MBR) remains attached to one daughter cell (Chen et al., 2013; Gromley et al., 2005). The fate of this MBR has recently attracted a lot of attention, as it might have crucial consequences for cell differentiation and tumorigenesis in mammalian cells (Chen et al., 2013). Indeed, cancer cells have been found to accumulate remnants, whereas normal cells do not (Ettinger et al., 2011; Kuo et al., 2011; Pohl and Jentsch, 2009). In addition, HeLa cells and mouse 134-4 hepatocarcinoma cells that accumulate high numbers of remnants display increased tumorigenicity, when compared to lower-accumulating cells of the same lines (Kuo et al., 2011). An increased number of MBRs in embryonic and adult fibroblasts is also associated with enhanced reprogramming efficiency and induced pluripotent stem cell (iPSC) colony formation (Kuo et al., 2011). Finally, in asymmetrical divisions in several tissues, the stem cell tends to ‘inherit’ the remnant, suggesting that remnants might have a role in maintaining stem cell status (Kuo et al., 2011). Conversely, differentiating cells have been found to preferentially discard remnants (Ettinger et al., 2011). Overall, the accumulation of remnants is associated with enhanced proliferation, whereas their efficient disposal is associated with normal cell differentiation. It thus appears important to understand how remnant disposal is regulated in normal and cancer cells.

To date, two pathways for remnant disposal have been described in mammalian cells. In the autophagy pathway (Fig. 5B, top row), the remnant is generated by a single ESCRT-dependent cut on one side of the midbody, and then retracts into the cytoplasm of one daughter cell (Chen et al., 2013). The cytoplasmic remnant is later wrapped in LC3-positive membranes, sequestered within autophagosomes and targeted to lysosomes for degradation (Ettinger et al., 2011; Kuo et al., 2011; Pohl and Jentsch, 2009). A second pathway has been described for neuroepithelial cells upon apical progenitor differentiation in vivo and for Neuro-2a and NS-5 cells in culture, wherein the remnant is released into the medium (Dubreuil et al., 2007; Ettinger et al., 2011; Marzesco et al., 2005). In this scenario, the bridge has to be cut twice through the assembly of the ESCRT machinery on both sides of the midbody. It is not yet clear to what extent cells use one pathway or another (Chen et al., 2013), but cancer cells have been suggested to accumulate remnants because of a failure to degrade them by autophagy (Kuo et al., 2011).

Because of the fundamental consequences that MBR retention or release could have on cell proliferation versus differentiation, we investigated in detail the fate of the MBR in multiple mammalian cell types. Surprisingly, we demonstrate that the great majority of remnants in a variety of cancer and immortalized cells, as well as in muscle stem cells, are actually released by sequential abscissions on both sides of the midbody, and do not retract into the cytoplasm. However, free MBRs remain associated with the cell surface for several hours, before being engulfed and then degraded in lysosomes. These observations lead to a novel model of how MBRs are inherited in mammalian cells and how this structure might regulate cell proliferation versus differentiation.

In many cultured cell lines (e.g. HeLa, MEF, U2OS), the midbody remnant is thought to remain continuous with one daughter cell after abscission and then to retract inside the cell for degradation by autophagy (Chen et al., 2013; Ettinger et al., 2011; Kuo et al., 2011; Pohl and Jentsch, 2009). Time-lapse microscopy in HeLa cells expressing mCherry fluorescent protein (mCherryFP)-tagged α-tubulin allowed us to follow the behavior of the midbody remnant, which appeared as a dense structure retaining prominent tubulin staining after microtubule severing and abscission (Fig. 1A, arrowheads), as reported previously (Elia et al., 2011; Guizetti et al., 2011; Lafaurie-Janvore et al., 2013; Schiel et al., 2012). Surprisingly, we observed that the MBR exhibited long-range movement after abscission. This striking displacement of the MBR could either result from rapid transport within the cytoplasm of the daughter cell or, more likely, from the movement of the MBR over the cell surface. Supporting the latter hypothesis, phase contrast images suggested that the MBR was in focus above the planes of the nucleus and cytoplasm (Fig. 1A, time-points 1:50 and 2:10).

To determine whether MBRs were indeed at the cell surface, we used correlative time-lapse light and scanning electron microscopy (SEM) to localize the midbody. When cells were fixed before abscission (Fig. 1B), the intercellular bridge connecting the sister cells was intact and displayed a membrane bulge at the exact position of the midbody observed by phase contrast imaging. Of note, the SEM protocol that we optimized preserved the bridge and midbody integrity and revealed high-resolution details in three dimensions. In particular, the membrane bulge around the midbody ring was striking (Fig. 1B, arrowheads) and is consistent with previously published serial electron microscopy sections through the bridge (Mullins and Biesele, 1977). Importantly, out of 23 cells fixed after abscission (3.1±1.3 h post-abscission; mean time±s.d.), the vast majority (20/23) displayed remnants that were liberated by two abscission events (both ends of the remnant were visible and sealed) and found at the cell surface (Fig. 1C). Remnants measured 4.6 µm in length on average, and varied from 0.6 to 13.5 µm (Fig. 1D). Midbodies retained the same morphology following abscission, and their diameter did not differ significantly in bridges and remnants (mean diameter for bridges, 1.6 µm; mean for remnants, 1.5 µm; two-tailed Student's t-test; P = 0.186). Out of the 23 MBRs analyzed by correlative SEM, only one was on the surface and still clearly connected on one side to a daughter cell, suggesting that a second abscission event follows the first within a short time.

We conclude that intercellular bridges in HeLa cells are cut on both sides of the midbody. This liberates a midbody remnant that is associated with and mobile at the cell surface of one of the daughter cells. This finding was confirmed in a variety of cell lines – mouse embryonic fibroblasts (MEFs), Madin-Darby canine kidney (MDCK), human kidney (HK)-2, U2OS (an osteosarcoma line), C2C12 (a mouse myoblast line) and RPE-1 (human retinal pigment epithelial cells) – as described below.

We hypothesized that MBRs were associated with the cell surface after abscission through strong interactions. Indeed, quantification of remnants in fixed cells using the ring marker MKLP1 (also known as KIF23) (Ettinger et al., 2011; Gromley et al., 2005; Joseph et al., 2012; Makyio et al., 2012; Steigemann et al., 2009) revealed that remnants are not removed even after extensively rinsing the cells (Fig. 2A). By contrast, treating the cells with trypsin-EDTA and immediately replating in normal medium resulted in the loss of ∼30–50% of remnants in HeLa, RPE-1, U2OS, HK-2, MDCK, C2C12 and MEF cells (Fig. 2B). This observation was not restricted to immortalized or cancer cell lines, because we also found that 68% of the MBRs were released from the cell surface by trypsin-EDTA in skeletal muscle stem (satellite) cells isolated from mouse muscles (Fig. 2B). In all cell lines analyzed, MBRs were released in similar proportions when cells were detached by EDTA treatment before replating (Fig. 2C). Of note, detached MBRs could be separated from cells by centrifugation and were positive for acetylated tubulin, MKLP1 and plasma membrane markers (Fig. 2D). Thus, in these cell types, a third to a half of remnants were resting on the cell surface, which likely involves interactions with a protein receptor that is dependent on Ca²⁺/Mg²⁺. As expected, Z-stack reconstruction frequently (73.3% of the cases) permitted unambiguous localization of the MBR either at the surface or inside the cell (n = 270 MBRs analyzed; Fig. 2E). Importantly, the results in Fig. 2B obtained in asynchronous cell populations also suggest that MBRs stay at the surface for several hours. Indeed, if 20% of cells with a 24-h cell cycle have an MBR, 50% of which is associated with the cell surface as in HeLa cells, it implies that MBRs remain at the surface for ∼4.8 h (2×20%×50%×24 h; multiplication by 2 corrects for the fact that the midbody is inherited by only one of the two daughter cells after abscission).

Time-lapse phase contrast microscopy using GFP-tagged Cep55 as a midbody marker (Bastos and Barr, 2010; Lee et al., 2008; Zhao et al., 2006) confirmed that MBRs displayed movement over the cell surface for extended periods of time. MBRs were found in focus on the cell surface in the majority of the divisions in multiple cell types; 98.4% of observations in HeLa (n = 64 cells), 100% of observations in U2OS (n = 55), 97.9% of observations in MEF (n = 48) and 98.3% of observations in MDCK (n = 115). Moreover, MBRs showed long-range displacement (over more than half the cell surface) in 78.1% of cell divisions in HeLa, 92.7% in U2OS; 85.4% in MEF and 96.5% in MDCK (Fig. 2F; supplementary material Fig. S1). Consistent with the estimation detailed above based on the results from Fig. 2B, quantification using phase contrast time-lapse microscopy revealed that the MBRs moved over the surface of HeLa cells for 4.3 h±2.5 (mean±s.d.; n = 44 cells) following abscission. Of note, in the example presented in Fig. 2F, the MBR moved from one daughter cell to the other. This unambiguously demonstrates that the MBR was cut twice and that it was present on the cell surface (see also supplementary material Movie 1; Fig. S1).

Interestingly, high-magnification time-lapse microscopy revealed that the MBR eventually disappeared from the phase contrast channel. This was accompanied by the remodeling of the plasma membrane at the MBR position, suggesting that it could have been internalized (Fig. 2G for U2OS cells; see also Fig. 4D for HeLa cells). This was demonstrated by the fact that the GFP–Cep55 fluorescent signal remained detectable after the disappearance of the MBR from the phase contrast channel (Fig. 2G, zooms), indicating that the MBR had not been released into the extracellular medium. This observation was confirmed in HK2, U2OS, MEF and MDCK cells (supplementary material Fig. S2). Although we observed rapid long-range movement of the MBR over the cell surface followed by internalization in most cases, we could also observe MBR release into the extracellular medium in a fraction of cases, depending on the cell type (e.g. 6/81 cases in MDCK cells, 5/22 in U2OS cells, but none in HeLa cells, n>200 cells). We conclude that, in the majority of the cells from multiple cell lines, remnants move over the cell surface for several hours after abscission, before being internalized.

After internalization, we noticed that the GFP–Cep55 fluorescent signal gradually decreased and eventually became undetectable (supplementary material Fig. S2 for HK2, MEF, MDCK and U2OS cells; Fig. S3A for HeLa cells). In HeLa cells, the GFP signal began to decrease at 2.9 h±2.0 (mean±s.d., n = 15 cells) after MBR disappearance from the surface.

Disappearance of the GFP signal was hypothesized to be due to acidification of the compartment containing the remnant, because GFP fluorescence is rapidly quenched at acidic pH (Kneen et al., 1998). In accordance with this idea, mCherryFP-tagged α-tubulin signal intensity was unaffected (supplementary material Fig. S3A), because this fluorescent protein is not sensitive to pH. As expected, the decrease in GFP–Cep55 fluorescent signal was concomitant with the recruitment of LysoTracker-labeled membrane around the midbody (Fig. 3A). We confirmed that, among all detected MBRs, ∼20% were indeed associated with Lamp1-positive lysosomal compartments (Fig. 3B,C), consistent with previous reports showing that a fraction of MBRs are associated with lysosomes. These studies also established that inhibition of lysosomal activities increased the proportion of cells harboring an MBR (Ettinger et al., 2011; Kuo et al., 2011; Pohl and Jentsch, 2009). We reinvestigated this finding by quantitative analysis and modeling to determine whether lysosomal degradation is the only route for MBR decay following internalization. Inhibition of lysosomal function by ammonium chloride treatment for 16 h and 24 h resulted in a gradual accumulation of remnants (Fig. 3D), and increased the proportion of remnants within Lamp1-positive compartments to 75% (Fig. 3C). Moreover, the intensity of MKLP1 staining of midbodies in bafilomycin-treated cells was shifted and was significantly higher than in control cells (Fig. 3E; n>1800; P<10⁻⁴; Kolmogorov–Smirnov test), suggesting that remnants remained intact in the absence of lysosomal activity. We recently developed a simple analytical quantitative mathematical model that predicts the fraction of cells with an MBR in a dynamic asynchronous population (Crowell et al., 2013). Using this model and the measured proportion of MKLP1 remnants in fixed cells, we estimated that the lifetime of post-abscission MBRs is 11.0±1.3 h for HeLa cells, a parameter that is difficult to directly measure by live-cell imaging (Crowell et al., 2013). Importantly, simulation of the complete inhibition of remnant degradation for 16 h and 24 h predicts an MBR accumulation that closely matches the data that we observed experimentally after lysosomal inhibition (Fig. 3D, black versus red crosses). We thus conclude that MBRs are long-lived organelles and that lysosomal activity is necessary and sufficient for MBR clearance in HeLa cells.

We next analyzed how an MBR present at the cell surface could reach intracellular acidic endosomal-lysosomal compartments. Non-phagocytic cells can internalize large particles and bacteria through a process called engulfment (Flannagan et al., 2012). This process requires interactions between the particle and the cell surface through various receptors, membrane deformation and eventually internalization by dynamic cortical F-actin remodeling. The plasma membrane marker GFP–Rab35 (Chesneau et al., 2012; Kouranti et al., 2006) revealed that plasma membrane extensions closely encircled remnants (Fig. 4A). Ultrastructural analysis by correlative light and SEM revealed that plasma membrane extensions and sheets tightly interacted with remnants (Fig. 4B). Notably, F-actin-rich structures that strikingly resembled phagocytosis-like cups were also observed to encircle or partially cover remnants in fixed cells (Fig. 4C; supplementary material Movie 2). To analyze the dynamics of these structures, we labeled F-actin in living cells with the mCherryFP–LifeAct probe (Miserey-Lenkei et al., 2010). We observed that a transient burst of F-actin polymerization occurred in a ring-like manner around remnants, just before their disappearance in the phase contrast channel (Fig. 4D, red and green circles). To test whether F-actin was necessary for remnant engulfment, we used time-lapse microscopy to track the fate of the MBR in cells treated with the F-actin-depolymerizing drug latrunculin A. As described above, we could precisely monitor remnant engulfment by quantifying GFP–Cep55 fluorescence intensity over time. Of note, we focused on MBRs that were already formed at the time of drug treatment, because latrunculin A would inhibit furrow ingression and thus impede the generation of new remnants. In control cells treated with DMSO alone, the GFP intensity at MBRs rapidly declined as the remnants were internalized and targeted to lysosomal compartments (Fig. 4E, left panel), as described above. By marked contrast, GFP intensity at MBRs was constant even after long periods of time, and the remnants remained at the cell surface in latrunculin-A-treated cells (Fig. 4E, middle and right panels). We conclude that MBRs are engulfed into cells through a phagocytosis-like mechanism.

Interestingly, MBRs could be transferred to and internalized by non-sister cells, as observed in HeLa cells (Fig. 5A, left; supplementary material Movie 3). Indeed, we observed that MBRs that initially migrated over the cell surface of one of the two daughter cells (red, cell outlines in white) could be transferred to a third neighboring cell (not red, cell outlines in blue), before being engulfed as shown by the disappearance of the MBR marker GFP–Cep55 (Fig. 5A, right). This indicates that the processes of MBR interaction, engulfment and degradation are not cell-autonomous and that MBRs can be potentially transferred and engulfed by distant cells. Taken together, these experiments demonstrate that free remnants generated by sequential abscission events are eventually engulfed through an F-actin-dependent mechanism and then targeted to lysosomes for degradation.

In mammalian cells, MBRs are thought to be either released into the extracellular medium or retracted into one of the two daughter cells, where they can be degraded by autophagy (Ettinger et al., 2011; Kuo et al., 2011; Pohl and Jentsch, 2009). Here, we describe a major alternative pathway observed in a variety of proliferating immortalized and cancer cell lines, and in skeletal muscle stem cells (Fig. 5B). Time-lapse light microscopy combined with correlative SEM revealed that, in the great majority of the cell divisions, the intercellular bridge is cleaved on both sides of the midbody, generating an MBR that is physically separated from the cell cytoplasm and moves for long distances over the cell surface (Fig. 5B, middle row). These two cleavage events are likely triggered by the local assembly of ESCRT-III helices ∼1 µm from the center of the midbody (Elia et al., 2011; Guizetti et al., 2011), consistent with the MBR dimensions observed by SEM (Fig. 1D). Because the ESCRT assembly occurs initially on one side of the midbody (Elia et al., 2011; Guizetti et al., 2011; Schiel et al., 2012), the first cut generates a nearly half-bridge containing the midbody that moves towards the daughter cell opposite to the first abscission site. Our correlative SEM observations indicate that 1 out of 23 midbodies was cut only once, whereas 20 were cut twice (hindered visibility of the ends prevented determination of whether the remaining two were cut once or twice). This result agrees with a recently published report demonstrating that a second abscission event is rapidly triggered by a release of tension in the bridge caused by the first abscission event (Lafaurie-Janvore et al., 2013). Thus, bridge retraction into the cytoplasm (Fig. 5B, top row) must be an uncommon event, which might be observed only if the second cut is delayed extensively. It has been reported that MBRs were found within LC3-associated autophagosomal vacuoles, consistent with the hypothesis that MBRs are degraded by autophagy after bridge retraction into the cytoplasm (Kuo et al., 2011; Pohl and Jentsch, 2009). Using GFP–LC3 or anti-LC3 antibodies, we observed MBRs within LC3-positive vacuoles only in few instances; 0.36% (n = 2794 MBRs) in HeLa cells, 0.53% (n = 378 MBRs) in RPE-1 cells, 0% (n = 490 MBRs) in U2OS cells and 0% (n = 582 MBRs) in MEF cells. However, as previously reported (Pohl and Jentsch, 2009), we consistently observed a faint uniform staining of the microtubule-binding protein LC3 in cytokinetic bridges before the abscission (and thus not directly related to autophagy), presumably because there is a high concentration of ubiquitylated proteins and microtubules at this location (Isakson et al., 2013). Thus, MBRs within LC3-positive vacuoles correspond to either rare or extremely transient events (estimated to last 2–3 min, based on the 11-h lifetime of an MBR in HeLa cells and the frequency of LC3-positive MBRs – 0.36%). Interestingly, there is accumulating evidence that the autophagy and phagocytic pathways can share common molecular components, such as Atg7, Atg6 and LC3 in the case of LC3-associated phagocytosis (LAP; Florey and Overholtzer, 2012). This might explain why the inactivation of the autophagy pathway by Atg7 knockout or depletion leads to a moderate increase in the number of MBRs in the cell population (Kuo et al., 2011; Pohl and Jentsch, 2009). In any case, both the autophagy and the phagocytic pathways merge and terminate in lysosomal compartments, explaining why inhibition of lysosomal activities using various means consistently leads to an increase in MBR lifetime (Ettinger et al., 2011; Kuo et al., 2011; Pohl and Jentsch, 2009; Fig. 3D). Consistent with MBR following a phagocytosis-like pathway and with previous analysis in Caenorhabditis elegans (Chai et al., 2012; Ou et al., 2014), we further observed that a fraction of the MBRs was associated with the early endosomal marker Rab5 and with the late endosomal/lysosomal marker Rab7. In addition, 18% of the MBRs were found in lysosomes that were positive for both Rab7 and Lamp1 (supplementary material Fig. S3B,C).

In summary, our data indicate that, in the multiple mammalian cell lines that we investigated, the vast majority of MBRs are cut twice and are either released into the medium (7% in MDCK cells, 22% in U2OS cells but 0% in HeLa cells) or engulfed through a phagocytosis-like mechanism. In HeLa cells, MBRs remain at the cell surface for ∼4 h and are then engulfed and enter into acidic lysosomal compartments after 2 h, where they are fully degraded after 4 h. Thus, MBRs are long-lived organelles that persist for most of the following cell cycle before being degraded. Our mathematical modeling further indicates that lysosomal activity fully accounts for the degradation of the MBRs, suggesting that other intracellular degradative mechanisms are likely not involved.

Extensive washes and EDTA treatment revealed that MBRs are tightly bound to the cell surface through a putative Ca²⁺/Mg²⁺-dependent receptor. We presently do not know the nature of this receptor, but we could occasionally detect clear β1-integrin staining around MBRs at the cell surface (data not shown). This is consistent with the occurence of integrin-dependent phagocytosis, which has been implicated in the clearance of particles, microorganisms and apoptotic cells (Dupuy and Caron, 2008). However, MBRs were not released from the cell surface after β1-integrin depletion, indicating either that it is not part of the MBR receptor or that other redundant receptors exist. Interestingly, we did not detect exposed phosphatidylserine on MBRs (by extracellular annexin-V staining on fixed HeLa cells without permeabilization; n = 200 MBRs). This indicates that either mammalian MBRs do not expose phosphatidylserine or that, as reported in C. elegans (Chai et al., 2012) phosphatidyserine is exposed very transiently (likely for <10 min). The lipidome and the proteome (after Triton X-100 extraction) of intercellular bridges revealed that they contained several hundred proteins and specific lipids (Atilla-Gokcumen et al., 2014; Skop et al., 2004). The comprehensive composition of transmembrane and exposed proteins at the bridge surface is currently unknown, thus future work will be needed to identify the proteins or lipids recognized on the MBRs.

Although we focused our study on proliferating mammalian cell lines, it has been established that MBRs are extensively released into the extracellular medium from differentiating cells, such as Neuro-2a and NS-5 cells (Ettinger et al., 2011). A simple explanation is that the MBR receptors either are not expressed or are expressed at a lower level when cells enter a differentiation program. Remarkably, recent observations indicate that the MBR receptors might not be ubiquitously expressed in vivo (Ou et al., 2014; Singh and Pohl, 2014). During the first two divisions of C. elegans embryos, it has been shown that the MBRs from the AB cell do not associate with the daughter cells ABa or ABp, but with the non-sister cell EMS. Similarly, in male germline, the MBRs produced from the spermatogonial divisions associate with non-sister cyst cells (Salzmann et al., 2014). Thus, MBRs likely display particular signals that are recognized by specific cell receptors in a wide range of organisms.

In the present study, we reveal that MBR engulfment is a feature shared by multiple mammalian cell types. Similarly, it has been reported that the four MBRs of the Q neuroblast lineage are phagocytosed by macrophages in vivo in C. elegans larvae (Chai et al., 2012). By contrast, it has been suggested that the MBR of the first cell division in C. elegans embryos retracts and is released into the posterior cell following an ESCRT-dependent abscission step (Green et al., 2013). However, very recent observations indicate that the MBR is actually engulfed after abscission (Ou et al., 2014; Singh and Pohl, 2014). MBR engulfment is also presumably observed in specific divisions of male germline cells in Drosophila (Salzmann et al., 2014). Taken together, these data suggest that, rather than being the exception, MBR engulfment is a distinct and evolutionarily conserved phenomenon. Interestingly, MBR engulfment in C. elegans relies on the conserved CED-1 and CED-2 pathways (Chai et al., 2012; Ou et al., 2014). It is tempting to speculate that this is also the case in Drosophila and mammals, and that the observed actin remodeling leading to MBR engulfment (Fig. 4D,E) thus depends on Rac GTPases.

One possibility is that MBR engulfment is fortuitous and has no biological function. Alternatively, engulfed MBRs might provide building blocks for the cells, particularly when they are stressed. However, this is unlikely, because MBRs accumulate when cells are incubated in an amino-acid-free culture medium (supplementary material Fig. S4A). Interestingly, the observed increase in the number of MBRs corresponded to MBRs at the cell surface, because they were released by EDTA treatment (supplementary material Fig. S4B). These experiments thus indicate that the capacity for MBR engulfment in a given cell type depends on the physiological state of the cell and can be modulated by external signals. Perhaps more interestingly, MBRs might have signaling functions, either once internalized by phagocytosis (as is the case for certain membrane receptors in so-called ‘signaling endosomes’; Scita and Di Fiore, 2010) or at the cell surface. The fact that MBRs display a stereotypical behavior and associate either with the stem versus differentiating cell or with specific non-daughter cells (Ou et al., 2014; Salzmann et al., 2014; Singh and Pohl, 2014) is remarkable and might indicate a specific role. An intriguing possibility is that the MBR might signal from the surface to the cell that will engulf it. Indeed, quantitative analysis of the number of MBRs found at the cell surface at any time in an asynchronous cell population indicates that MBRs spend a third to half of their lifetime at the cell surface in a variety of immortalized or cancer cell lines (C2C12, HeLa, HK-2, MEF, RPE-1, U2OS) and primary satellite stem cells. Therefore, MBRs could potentially cluster receptors and signal to the recipient cell for several hours (4–5 h in the example of HeLa cells) before being engulfed, which, in this hypothesis, would serve to terminate the signal. If true, this might explain why no overall developmental defects have been observed in C. elegans mutants in which MBR internalization has been impaired (Chai et al., 2012; Ou et al., 2014), whereas full destruction of MBRs by laser ablation impairs dorso-ventral axis formation (Singh and Pohl, 2014). Cell cycle progression seems to be delayed after laser ablation of MBRs, suggesting that either this treatment alters additional cell structures or the MBR also influences the timing of cell division. Additional studies will thus help to further delineate the physiological functions of the MBRs in C. elegans development.

It has been proposed that MBRs could influence differentiation versus proliferation in mammalian cells (Ettinger et al., 2011; Kuo et al., 2011; Pohl and Jentsch, 2009). However, this exciting possibility might be cell-type dependent, and the initial thought that MBRs would inhibit differentiation and/or activate proliferation is likely an oversimplification, as recently suggested based on data from Drosophila germlines (Salzmann et al., 2014). Nevertheless, the intriguing observation that MBRs are present in the ventricular fluid during mouse brain development and the potential engulfment figures seen in vivo in fixed samples (Dubreuil et al., 2007; Ettinger et al., 2011; Marzesco et al., 2005) raises the exciting possibility that MBRs could serve as vehicles for long-distance intercellular signaling, provided that the appropriate receptor is expressed by the recipient cell. The transfer of the MBRs between non-sister cells experimentally indicates that this is possible, at least in cells in culture (Fig. 5A,B, last row; supplementary material Movie 3). Overall, our study changes our vision of how MBRs are inherited and degraded in mammalian cells, and suggests mechanistic insights into how remnants might signal over long distances between non-sister cells.

HeLa, U2OS, RPE-1, HK-2 and MEF cells were cultivated in Dulbecco's Modified Eagle Medium (DMEM) GlutaMax (31966; Gibco, Invitrogen Life Technologies) supplemented with 10% fetal bovine serum (FBS) (PAN Biotech) and 1% penicillin-streptomycin (15140; Gibco, Invitrogen Life Technologies) (except for MEF cells, which were cultured without antibiotics). C2C12 were cultivated in DMEM with 20% FBS and 1% penicillin-streptomycin. MDCK cells were cultivated in Modified Eagle alpha Medium (22571; Gibco, Invitrogen, Life Technologies) supplemented with 10% FBS, 2 mM L-glutamine (25030; Gibco) and 1% penicillin-streptomycin. Cells were grown at 37°C under 5% CO₂. For experimentation, cells were detached from flasks with 0.05% trypsin diluted in 0.02% EDTA (25300; Gibco, Invitrogen Life Technologies) and plated at 1×10⁴ or 2×10⁴ cells/well on glass coverslips in 24-well plates (92024; TPP).

Skeletal muscle stem (satellite) cells were isolated and cultured as described previously (Gayraud-Morel et al., 2012). Briefly, cells were prepared from Tg:Pax7-nGFP mouse hindlimb and forelimb muscles. After removal of the major tendons, nerves and adipose tissue, muscle tissue was minced with scissors and digested. Cells were then centrifuged at 515 g for 15 min at 4°C. After several washes, isolation of cells was performed on a FACSAria (BD Biosciences). Cells were finally plated on a 96-well plate (655892; Greiner Bio-One) at 1000 cells per well and cultured in 1∶1 DMEM∶MCDB (Sigma-Aldrich), 20% fetal calf serum (22477K; Gibco, Invitrogen, Life Technologies) at 37°C under 5% CO₂ and 3% O₂. Cells were fixed at 96 h after plating.

Plasmids encoding GFP–Rab35 have been described previously (Chesneau et al., 2012; Dambournet et al., 2011), as were those encoding LifeAct–mCherryFP (Miserey-Lenkei et al., 2010) and mCherryFP–α-tubulin (Connell et al., 2009; Guizetti et al., 2011; Lafaurie-Janvore et al., 2013). Plasmid encoding GFP–Cep55 was a kind gift from E. Laplantine and P. Genin.

Cells were transfected for 24 h to 48 h before live imaging or fixation using X-trem9 (HeLa, U2OS and HK-2 cells) (Roche), Lipofectamine 2000 (MEF cells) (Invitrogen, Life Technologies) or Amaxa nucleofection (MDCK cells) (Lonza).

Mouse anti-acetylated tubulin (611B1; Sigma-Aldrich) and rabbit anti-Aurora-B (AP7240a; AbGENT) were used at 1∶1000; mouse anti-Lamp-1 antibody (555798, BD Pharmingen) was used at 1∶100. Alexa-Fluor-488–phalloidin (Molecular Probes) was used to label F-actin in fixed samples as described previously (Miserey-Lenkei et al., 2010). Alexa-Fluor-568–WGA (wheat germ agglutinin) was used to label the MBR membrane (45-min incubation at 1∶200). Two different rabbit anti-MKLP1 antibodies were used. In order to label human cells, we used anti-MKLP1-N-19 (sc-867; Santa Cruz) at 1∶300. For mouse C2C12, MEF and satellite cells, we used an anti-KIF23 antibody, generously provided by D. Fesquet, diluted 1∶1000. Cells were fixed in 10% trichloroacetic acid (T8657; Sigma Aldrich) in unsupplemented medium for 20 min on ice, washed in Dulbecco's phosphate-buffered saline (DPBS) (14040; Gibco, Invitrogen, Life Technologies) and permeabilized in DPBS (D1408; Sigma-Aldrich) containing 0.2% bovine serum albumin (BSA; A7030; Sigma Aldrich) and 0.1–0.5% Triton X-100 for 3 to 5 min. The blocking steps (20 min), primary antibody incubation (1 h), rinses (3× 5 min) and secondary antibody incubation (30–45 min) were performed in DPBS-BSA. Dylight Alexa-Fluor-488- and Cy3-conjugated secondary antibodies (Jackson Laboratories) were diluted 1∶500. DAPI staining (1 mg/ml, Serva) was performed for 5 min following secondary antibody incubation. Cells were mounted in Mowiol.

Immortalized cancer cells were plated on 12-mm coverslips in 24-well plates for 48 h before experiments. Satellite cells were treated 96 h after isolation.

Following a brief wash with DPBS, cells were detached either with 0.05% trypsin diluted in 0.02% EDTA or with 0.02% EDTA alone. Finally, cells were replated on fibronectin-coated coverslips (F1141; Sigma Aldrich) for a maximum of 2 h before fixation. Control cells were not detached but were rinsed as well as treated cells. For each cell line, at least three independent experiments were performed with a minimum of two replicates per experiment. To demonstrate that rinses do not induce MBR release, cells were rinsed 0, 1, 2, 3, 4 or 5 times using DPBS right before fixation (three independent experiments, three replicates per experiment).

Cells were incubated in 0.02% EDTA for 20 min at 37°C in order to detach MBRs from the cell surface. The supernatant was collected and briefly centrifuged to remove cells (5 min at 50 g). Finally, MBRs were sedimented by centrifugation for 30 min at 1200 g, on fibronectin- and poly-l-lysine-coated 96-well plates.

HeLa cells were treated for 16 h or 24 h with 10 mM NH₄Cl or vehicle (DPBS) before fixation. The control condition presented in Fig. 3D corresponds to the 16-h treatment but was similar to the control condition of the 24-h treatment. Three independent experiments were performed with three replicates per experiment.

HeLa cells were treated for 24 h with 200 nM bafilomycin or solvent (DMSO) before fixation. Analysis of MKLP1 staining intensity of remnants is described below.

Cells plated on 35-mm glass-bottomed dishes (MatTek Advanced TC) were treated with 0.5 µM latrunculin A or solvent (DMSO) at 30 min before the start of time-lapse microscopy. Cells that had already successfully ingressed cytokinetic furrows and formed a bridge were selected for imaging. GFP–Cep55 fluorescence at the midbody was monitored by automated tracking of the midbody in time series using an algorithm developed in ImageJ. Mean intensity over time was plotted using Matlab.

Cells were rinsed three times and incubated for 4 h or 7 h either in normal culture medium (control conditions) or in EBSS medium supplemented with Ca²⁺ and Mg²⁺ (5.3 mM KCl, 26.2 mM NaHCO₃, 117.2 mM NaCl, 1.0 mM NaH₂PO₄•H₂O, 1.98 mM CaCl₂•2H₂O, 0.8 mM MgSO₄•7H₂O and 5.6 mM glucose) (starvation conditions). After incubation, cells were fixed and processed for immunofluorescence. To analyze the proportion of MBRs at the cell surface, EDTA treatments were performed 2 h before fixation, as described above.

Cells were plated on 35-mm glass-bottomed dishes (MatTek Advanced TC). They were maintained at 37°C under 5% CO₂ in an imaging chamber mounted on a Nikon Eclipse Ti inverted microscope equipped with Nikon 20× 0.45 NA Plan Fluor ELWD and 60× 1.4 NA Plan Apochromat objectives, a motorized stage and an HQ2 cooled CCD camera (Roper Scientific). The microscope was piloted using MetaMorph (MDS Analytical Technologies). ‘Perfect Focus System’ (PFS) was used to maintain the focus during time series of a 24–72-h duration with a time interval of 5–10 min. Z-stacks of 15 slices (0.5-mm Z-interval) were acquired when the 60× objective was used. For LysoTracker experiments, cells were incubated with LysoTracker Red DND-99 (L-7528; Life technologies) following the manufacturer's instructions, and no rinses were performed, in order to compensate for the rapid photobleaching of the dye. Time-lapse microscopy was started at 10 min after adding the dye.

Cells were loaded onto cell-locator glass-bottomed dishes (MatTek Corporation) coated with poly-l-lysine. Cells were fixed in a solution of 4% paraformaldehyde (PFA) and 0.1% glutaraldehyde and prepared for correlative light scanning electron microscopy. Specific areas were imaged and localized with high resolution on the cell-locator glass-bottomed dishes by using the microscope described above. For subsequent SEM analysis, the same cells were refixed with 2.5% glutaraldehyde in 0.2 M cacodylate buffer (pH 7.2) overnight at 4°C, then washed for 5 min three times in 0.2 M cacodylate buffer (pH 7.2), post-fixed for 1 h in 1% osmium and rinsed with distilled water. Cells were dehydrated through a graded series of ethanol followed by critical-point drying with CO₂. Dried specimens were sputter-coated twice with gold-palladium with a gun ionic evaporator PEC 682. The coordinates of the correlative cells imaged with fluorescent microscopy were recovered in a JEOL JSM 6700F field emission scanning electron microscope operating at 7 kV.

Cells fixed on coverslips and stained with DAPI, anti-MKLP1 and anti-acetylated tubulin or anti-Aurora-B were scanned using an automated journal in MetaMorph and the motorized Nikon Eclipse Ti described above. Imaging of satellite cells was directly performed in 96-well plates using a modified version of this journal. A least 500 cells per condition were analyzed. In NH₄Cl experiments, cells were imaged and analyzed using a Nikon Eclipse Ti inverted microscope equipped with 60× and 100× 1.4 NA Plan Apochromat objectives and an HQ2 cooled CCD camera (Roper Scientific) piloted by MetaView. For LifeAct staining shown in Fig. 4D and Z-profiles shown in Fig. 2E, images were deconvolved using Huygens software (Scientific Volume Imaging) with the microscope PFS (10 iterations, threshold 20).

Quantification was performed either manually or automatically under visual inspection, based on MKLP1 and acetylated tubulin staining using random samples of >500 cells. For automatic quantification, pre-processing, segmentation and measurements were performed using an algorithm developed in ImageJ, and object classification was achieved using a program written in TCL. The same algorithm permitted the measurement of MKLP1 staining intensity at remnants (data presented in Fig. 3E).

GFP–Cep55 fluorescence at the midbody was monitored by automated tracking of the midbody in time series using an algorithm developed in ImageJ. Mean intensity over time was plotted using Matlab, and the time at which GFP intensity dropped was estimated by visual inspection.

HeLa, U2OS, MDCK and MEF cells were transfected with mCherryFP–α-tubulin and GFP–Cep55, and imaged for 24–48 h at 1 frame every 5–10 min. For each transfected and dividing cell, the movement of the MBR was followed and characterized as ‘long-range’ if it covered a distance superior to half the size of the cell before being internalized or released. The MBR was considered as internalized if it stayed for several minutes at the same location before disappearing from the phase contrast channel and then gradually from the GFP channel. It was considered as released if it suddenly disappeared simultaneously from the two channels.

Remnants whose two ends were clearly visible and sealed were measured using standard line tools in ImageJ.

Statistical analysis of experimental data was performed in R 2.15.1 GUI 1.52 64-bit or GraphPad Prism. Model calculations were executed using Physics Analysis Workstation (PAW) or Matlab (The MathWorks, Natick, MA). Graphs were produced using either GraphPad Prism or Matlab.

Error bars represent s.d. The significance of differences between various conditions was calculated using Student's t-test or the Kolmogorov–Smirnov test. *P<0.05; **P<0.01; ***P<0.001; ****P<0.0001.

We thank Didier Fresquet (Centre de Recherche de Biochimie Macromoléculaire CNRS, Montpellier, France), Sandrine Vitry, Emmanuel Laplantine and Pierre Genin (Institut Pasteur, Paris, France) for reagents and plasmids; Adeline Mallet for training in correlative light scanning and electron microscopy; Marie Delattre, Stéphanie Miserey-Lenkei and the Echard laboratory members for critical reading and discussion. We thank the imaging facility Imagopole Institut Pasteur and Kerstin Klinkert for help with image processing.