Mammalian erythroblast enucleation requires PI3K-dependent cell polarization

In the last step of erythropoiesis, mammalian erythroblasts undergo enucleation, a process that is crucial for the formation of mature functional red blood cells. During enucleation, the erythroblast extrudes its nucleus tightly apposed to the plasma membrane, forming a reticulocyte (Ihle and Gilliland, 2007; Koury et al., 2002; Richmond et al., 2005). Pioneering studies using electron microscopy revealed that at the earliest stage of enucleation the erythroblast nucleus becomes located close to the cell membrane away from the center of the cell (Simpson and Kling, 1967; Skutelsky and Danon, 1967) and that a cytokinetic-like furrow is formed in the region between the extruded nucleus and the incipient reticulocyte (Koury et al., 1989; Skutelsky and Danon, 1967). Actin filaments (F-actin) accumulate in the cytokinetic-like furrow (Ji et al., 2008; Koury et al., 1989) and disruption of F-actin (Ji et al., 2008; Koury et al., 1989; Yoshida et al., 2005) or depletion of mDia2, a regulator of actin polymerization (Ji et al., 2008), blocked enucleation, suggesting that actin-based forces drive nuclear extrusion. However, many questions remain unanswered concerning the process of erythroblast enucleation. In particular, little is known how an asymmetry is established within the erythroblast (i.e. how the nucleus becomes localized to one side of the cell and the cytoplasm to the other), although this polarized state appears to be important for enucleation. Moreover, the detailed organization of actin and microtubules in polarized erythroblasts is unknown.

Phosphoinositide 3-kinase (PI3K) is well known as a central regulator of chemotaxis. In migrating Dictyostelium discoideum, neutrophils and fibroblasts, the PI3K products PtdIns(3,4)P₃ and PtdIns(3,4,5)P₃ accumulate locally at the leading edge of the surface membrane and control cell polarization (Haugh et al., 2000; Parent et al., 1998; Servant et al., 2000). Although involvement of PI3K in the early stages of Epo (erythropoietin)-regulated differentiation of erythroid progenitors has been established (Ghaffari et al., 2006; Zhao et al., 2006), little is known about its role in the much later steps of enucleation.

We investigated how erythroblasts establish cell polarization and whether this polarization plays a role in expelling the nucleus from the cell. We used a powerful combination of an in vitro cell culture system that mimics normal terminal erythroid proliferation, differentiation and enucleation (Ji et al., 2008), combined with several microscopic imaging techniques. Our results show that proper enucleation requires establishment and maintenance of cell polarization mediated by PI3K in a manner similar to that seen in migrating cells.

We first wanted to determine when the terminal erythroblast becomes polarized and how the nucleus is extruded from the erythroblast. To this end, we conducted a detailed microscopic analysis of the enucleation process using an in vitro cell culture system employing mouse fetal liver erythroblasts. Enucleation begins ~35 hours after stimulation of erythroid progenitors (Ji et al., 2008). This system employs normal primary erythroid cells and thus does not have the obvious abnormalities associated with virus-infected and/or transformed cell lines. Moreover, the time-course of erythroid differentiation in this system has been well established (Ji et al., 2008), providing us in a time-dependent manner with erythroblasts at different stages of differentiation.

We could follow cultured erythroblast cells after they underwent a final mitotic division (Fig. 1A, −4 minutes) and generated two late erythroblasts in which the nuclei were located at the center of each cell (Fig. 1A, 44 minutes). At ~3 hours after each was generated (210±8.64 minutes, n=8), these late erythroblasts underwent significant polarization, displacing the nucleus to one side of the cell (Fig. 1A, 115 and 153 minutes, arrows). Around 1 hour after cell polarization, enucleation initiated (74.7±8.64 minutes, n=8) and dynamic and random contractions occurred at the side of the cytoplasm opposite to the nucleus (n=16; Fig. 1B and supplementary material Movie 1). Strikingly, a part of the nucleus suddenly protruded from the cell, forming a bleb-like structure from a limited area of the cortex adjacent to the nucleus (3.35±0.64 μm diameter; Fig. 1B, arrows). Then the whole nucleus was quickly squeezed out (5.98±2.73 minutes), in the process undergoing extensive deformations into an hourglass-like shape (supplementary material Movie 1). Fixed late erythroblasts at different stages of enucleation (Fig. 1C) were morphologically similar to live cells (Fig. 1B).

Before late erythroblasts became polarized, F-actin formed small particles and patch-like structures distributed throughout the cytoplasm (Fig. 1D, Non-polarized). After cell polarization, F-actin was mostly restricted to the side of the cytoplasm away from the nucleus, and little was localized at the side with the nucleus (Fig. 1D, Polarized). While the nucleus was being expelled, a fraction of F-actin concentrated at the neck region of the bleb (Fig. 1D, Extruding nucleus, F-actin, arrowheads) where it was previously described as the cytokinetic-like furrow or contractile actin ring (Ji et al., 2008; Koury et al., 1989; Skutelsky and Danon, 1967).

Similar to F-actin, after late erythroblasts became polarized small particles of active myosin II (myosin II regulatory light chain phosphorylated at Ser19; pMRLC) (Matsumura et al., 1998) became restricted to the side of the cytoplasm away from the nucleus (Fig. 1D, pMRLC). A fraction of active myosin II also became localized to the neck region of the bleb (Fig. 1D, pMRLC, arrows).

Although previous studies have shown that F-actin is required for enucleation (Ji et al., 2008; Koury et al., 1989; Repasky and Eckert, 1981; Yoshida et al., 2005), it remained unknown whether enucleation is powered by contractions of asymmetrically localized actomyosin. To this end, we disrupted F-actin or inhibited myosin II function specifically in polarized late erythroblasts. We applied specific inhibitors of actin or myosin II to cells at 36 hours of culture, and polarized cells were then identified for time-lapse recording. In polarized cells treated with cytochalasin D, an agent that disrupts F-actin, dynamic contractions were quickly and completely suppressed and enucleation was blocked (Fig. 1E, Cytochalasin D).

Similarly, treatment of polarized cells with blebbistatin, an inhibitor of myosin II activity (Straight et al., 2003), at a concentration (100 μM) that completely inhibits cytokinesis (Mukhina et al., 2007), caused the suppression of cytoplasmic contractions and a failure of enucleation (Fig. 1E, Blebbistatin, top). Strikingly, even in cells already extruding the nucleus, following blebbistatin treatment the nucleus regressed into the cell (Fig. 1E, Blebbistatin, bottom). These results suggest that nuclear extrusion is achieved by pressure generated by contractions of asymmetrically localized actomyosin and thus that the establishment of cell polarization is crucial for enucleation. Our observations also indicate that completion of enucleation probably involves actomyosin-driven constriction of the cortical actin ring at the neck of the bleb-like protrusion.

How is cell polarization established and maintained during enucleation? We speculated that PI3K, which regulates cell polarization in migrating cells, might be involved in this process. To test this hypothesis, we first examined the localization of the PI3K products PtdIns(3,4)P₂ and PtdIns(3,4,5)P₃ by expressing the reporter (3)Akt(PH)–GFP (the trimerized PH domain of Akt tagged with green fluorescent protein) in late erythroblasts (Fig. 2A) (Luo et al., 2005).

In non-polarized cells, (3)Akt(PH)–GFP was localized throughout the entire plasma membrane (n=11; Fig. 2A, Non-polarized). Strikingly, when cells became polarized, (3)Akt(PH)–GFP became restricted in localization to the segment of plasma membrane facing the cytoplasm (n=8; Fig. 2A, Polarized); this localization pattern remained unchanged during nuclear extrusion (n=16; Fig. 2A, Extruding nucleus). Three-dimensional images of (3)Akt(PH)–GFP and mCherry, used as a cytoplasmic volume marker, showed that the asymmetric distribution of (3)Akt(PH)–GFP in polarized cells was not caused by volume effects in the cytoplasm because mCherry was uniformly distributed throughout the cell (Fig. 2B). In late erythroblasts, (3)Akt(PH^R25C)–GFP, a mutant that is unable to bind to PtdIns(3,4)P₂ or PtdIns(3,4,5)P₃ (Franke et al., 1997), failed to localize to the plasma membrane and was diffusely distributed in the cytoplasm, indicating that (3)Akt(PH)–GFP is a specific reporter for PI3K activity (supplementary material Fig. S1).

Next, we wanted to determine whether PI3K activity is required for cell polarization during enucleation. Because PI3K activity is involved in the early steps of Epo-triggered cell division and differentiation of erythroid progenitor cells (Ghaffari et al., 2006; Zhao et al., 2006), we needed to use a specific inhibitor of PI3K to block its activity only in late erythroblasts – more specifically during enucleation. To this end, we applied LY294002, a well-established specific inhibitor of PI3K, to late erythroblasts expressing (3)Akt(PH)–GFP. Strikingly, after ~1 hour of incubation the membrane association of (3)Akt(PH)–GFP became strongly diminished in 24 cells out of 28 (24/28) treated with LY294002 but not in the DMSO-treated control cells (0/17) (Fig. 2C), indicating that LY294002 effectively inhibits PI3K activity in late erythroblasts.

To determine whether PI3K inhibition affects the establishment of cell polarization, we applied LY294002 to cells at 31 hours in culture, where ~80% of late erythroblasts were not yet polarized (see supplementary material Fig. S2). After 1 hour of incubation, non-polarized cells were identified and monitored by time-lapse recording. The majority of control non-polarized cells underwent cell polarization within 3 hours (80%, n=40), whereas 75% of the non-polarized cells treated with LY294002 failed to become polarized during this time period (n=36; Fig. 3A and Fig. 4). The distance between the cellular and nuclear centroids in polarized control cells was 1.59±0.14 μm, whereas the maximum distance between these centroids was 0.62±0.09 μm in LY294002-treated cells that were not able to undergo cell polarization (Fig. 3B), suggesting that the nucleus remained localized in the cell center in LY294002-treated cells. F-actin was found localized throughout the cytoplasm in the LY294002-treated cells that remained non-polarized (n=23; Fig. 3C).

LY294002- and DMSO-treated (control) cells were also fixed in order to perform quantitative analysis on the progression of enucleation by counting the number of late erythroblasts at each stage of enucleation per total number of late erythroblasts and enucleated cells. We observed only a slight (although statistically significant) decrease in the percentage of polarized cells as well as enucleated cells in cells treated with LY294002 from 31 to 35 hours in culture, compared with control cells (polarized cells, 17.56±1.84% vs 24.48±2.67%; enucleated cells, 7.54±0.68% vs 14.24±1.80%; P<0.05, Fig. 3D). Similar results were obtained when late erythroblasts were treated with another PI3K inhibitor, wortmannin (supplementary material Fig. S3), suggesting that the defects in enucleation caused by LY294002 were indeed due to inhibition of PI3K activity and that PI3K inhibition did not completely block cell polarization. When cells were treated with LY294002 from 31 to 48 hours in culture, the decrease in polarization and enucleation was no longer statistically significant (Fig. 3E). These results suggest that PI3K inhibition caused a severe delay rather than arrest in the establishment of cell polarization.

To determine whether PI3K inhibition affects nuclear extrusion, LY294002 was applied to cells at 35 hours in culture and, after 1 hour of incubation, polarized cells were identified and monitored by time-lapse video microscopy. Strikingly, 76% of polarized cells treated with LY294002 exhibited a wide range of delay and/or arrest in nuclear extrusion (35/46), whereas in all control cells enucleation was completed within 25 minutes (46/46; Fig. 5A). Some cells completed nuclear extrusion after a delay (11/46), whereas others moved out of the field of view before the extrusion was completed or did not complete nuclear extrusion during time-lapse imaging (24/46, Fig. 4 and supplementary material Fig. S4).

Cells that showed a delay in nuclear extrusion typically exhibited one or both of the following phenotypes: (a) Nuclear extrusion was initiated but the nucleus regressed into the cell before the extrusion was completed; some cells repeated this ‘trial and error’ of nuclear extrusion several times. (b) Nuclear extrusion was initiated but failed to complete, with formation of an hourglass-shaped nucleus (Fig. 5A, supplementary material Movies 2 and 3).

In control cells, nuclear extrusion was completed, with a distance between the cellular and nuclear centroids of 6.11±0.40 μm (Fig. 5B, DMSO), whereas in LY294002-treated cells that showed a delay in nuclear extrusion the distance between these centroids was 1.59±0.32 μm and remained relatively unchanged until extrusion was completed (Fig. 5B, LY294002). Quantitative analysis revealed no significant decrease in the percentage of enucleated cells among cells treated with LY294002 from 35 to 39 hours in culture compared with control cells (Fig. 5C), suggesting that PI3K inhibition caused a delay rather than a complete arrest in nuclear extrusion. These results suggest that PI3K activity is important to maintain cell polarization. Taken together, our observations suggest that PI3K activity is required for proper enucleation by regulating the establishment and maintenance of cell polarization.

Next, we wanted to determine how PI3K activity is regulated during enucleation. In both migrating neutrophils (Xu et al., 2005) and macrophages during phagocytosis (Khandani et al., 2007), localized PI3K activity is dependent on microtubules. Thus, we speculated that microtubules might regulate PI3K activity during enucleation. Before testing this hypothesis, it was important to analyze the microtubule organization during enucleation because this had not been determined, especially in polarized late erythroblasts.

Immunofluorescence showed that, as previously described, before cells became polarized microtubule arrays were radially symmetric, forming a cage-like structure surrounding the nucleus (Fig. 6A) (Koury et al., 1989). Concomitant with cell polarization, microtubules formed an asymmetric array oriented toward the cortex on the side of the cytoplasm opposite to the nucleus (Fig. 6A, Polarized and Extruding nucleus). We also found that, during enucleation, one or two foci of GFP–γ-tubulin were localized together in the cytoplasm in close proximity to the nucleus at the side of the cytoplasm (n=6; Fig. 6B).

Similar localization patterns of endogenous γ-tubulin (n=21) and pericentrin (n=11) were observed in late erythroblasts (supplementary material Fig. S5A,B). Moreover, when late erythroblasts were stained for EB1, a protein that binds to plus-ends of growing microtubules (Mimori-Kiyosue et al., 2000), typical comet-like structures were observed regardless of the stage of enucleation (Fig. 6C, arrows), suggesting that microtubules are dynamic during enucleation. These results suggest that microtubule organization is dramatically rearranged into an asymmetric array with the centrosome when late erythroblasts become polarized.

To determine whether microtubules regulate PI3K activity during enucleation, late erythroblasts expressing (3)Akt (PH)–GFP were treated with nocodazole, a microtubule-depolymerization agent. Strikingly, in nocodazole-treated cells where microtubules were severely disrupted (3)Akt(PH)–GFP failed to associate with the plasma membrane and instead was diffusely localized in the cytoplasm (11/15). By contrast, (3)Akt(PH)–GFP was associated with the membrane in control cells (12/12) (Fig. 6D,E), suggesting that microtubules are important for maintaining PI3K activity.

Nocodazole treatment caused defects in enucleation similar to those seen in cells treated with LY294002. Live-cell analysis showed that when nocodazole was applied to cells at 32 hours in culture only 20% of non-polarized cells were able to undergo polarization within 3 hours (n=20; Fig. 6F, Fig. 4 and supplementary material Fig. S5C). F-actin was localized throughout the cytoplasm in nocodazole-treated cells that remained non-polarized (n=20; supplementary material Fig. S5D). When nocodazole was applied to cells at 36 hours in culture, 63% of polarized cells exhibited a severe delay in nuclear extrusion (n=35; Fig. 6G, Fig. 4, supplementary material Fig. S4, Fig. S5E and Movie 4), with phenotypes similar to those seen in LY294002-treated cells (Fig. 5A).

Quantitative analysis revealed only a slight decrease in the percentage of polarized and enucleated cells among cells treated with nocodazole from 32 to 36 hours in culture, compared with control cells (supplementary material Fig. S5F). It is noteworthy that treatment of erythroblasts with nocodazole from 32 to 48 hours in culture resulted in severe defects in nuclear and cellular morphology (data not shown). The decrease was not significant when cells were treated with nocodazole from 36 to 40 hours in culture (supplementary material Fig. S5G). These results suggest that microtubules are required for proper enucleation by regulating the establishment and maintenance of cell polarization. As detailed below, in large measure this resolves past controversies about whether or not microtubules play an essential role in enucleation (Keerthivasan et al., 2010; Koury et al., 1989; Sonoda et al., 1998). Unlike the crucial role of microtubules in the regulation of PI3K activity, PI3K inhibition caused no drastic defects in microtubule organization in late erythroblasts (Fig. 6H). Taken together, our observations suggest that microtubule-dependent local activation of PI3K is required for proper enucleation.

In migrating fibroblasts, reorientation of the microtubule-organizing center (MTOC) is the initial polarization event and occurs by the movement of the nucleus but not the MTOC (Gomes et al., 2005; Schmoranzer et al., 2009). Our live-cell analyses showed that when late erythroblasts became polarized the nucleus moved to one side of the cell (Fig. 3A,B), suggesting that cell polarization might be initiated in a way similar to that seen in migrating cells and that PI3K might regulate this process. To test this, cells treated with LY294002 from 31 to 35 hours in culture were fixed and stained for the nucleus and γ-tubulin (see supplementary material Fig. S6). Then, we measured the distance between the cellular centroid and the MTOC or the nuclear centroid (Fig. 7).

In control cells (Fig. 7, DMSO), the average distance between the MTOC and the cellular centroid in non-polarized cells (1.75±0.06 μm) was similar to that in polarized cells (1.71±0.08 μm), suggesting that the position of the MTOC remains unchanged during cell polarization. By contrast, as observed in live-cell analysis (Fig. 3A,B), the distance between the nuclear and cellular centroids was significantly increased when the cells became polarized (Fig. 7, DMSO). These results indicate that during the early stages of enucleation, MTOC reorientation occurs by the movement of the nucleus but not the MTOC.

In non-polarized cells, LY294002 treatment had no effect on the distances between the MTOC or the nuclear centroid and the cellular centroid (Fig. 7, Non-polarized, LY294002). In the few LY294002-treated cells that managed to become polarized, the distance between the nuclear and the cellular centroids was increased, whereas the position of the MTOC remained unchanged, similar to that seen in control DMSO-treated cells (Fig. 7, Polarized, LY294002). These results suggest that PI3K inhibition causes a delay in nuclear repositioning. Taken together, our observations suggest that PI3K regulates cell polarization through promoting the movement of the nucleus but not the MTOC during enucleation.

One of the long-standing questions in erythropoiesis is how the erythroblast extrudes its nucleus to form the reticulocyte that later becomes the mature red blood cell. In particular, little is known about how erythroblasts establish and maintain a polarized state during enucleation. The live-cell analyses we performed in this study have provided a deeper understanding of the mechanisms underlying the regulation of cell polarization and its importance in enucleation.

Early work suggested a mechanism of enucleation similar to that used in cytokinesis of mitotic cells, where local cortical contractions occur between the two daughter cells (Skutelsky and Danon, 1967; Koury et al., 1989). By contrast, our observations demonstrate that nuclear extrusion is driven by contractile forces generated by asymmetrically distributed cytoplasmic actomyosin in late erythroblasts. This finding reiterates the importance of establishment and maintenance of cell polarization during enucleation.

Although previous studies revealed that PI3K was required for early Epo-dependent stages of erythropoiesis (Ghaffari et al., 2006; Zhao et al., 2006), here we demonstrated a second and novel role for PI3K in the establishment and maintenance of cell polarization during enucleation. Moreover, our results also showed that PI3K promotes cell polarization by repositioning of the nucleus during enucleation. In migrating fibroblasts, cell polarization is established by the movement of the nucleus instead of the MTOC (Gomes et al., 2005). Our results indicate that this is also the case in late erythroblasts during enucleation. A possible mechanism that regulates the repositioning of the nucleus by PI3K is discussed below.

The roles of microtubules in enucleation have been controversial (Keerthivasan et al., 2010; Koury et al., 1989; Sonoda et al., 1998). We observed different effects of microtubule disruption agents on enucleation when cells were treated at different stages of erythropoiesis and/or for different time periods of incubation. Imaging and quantitative analysis of live and fixed cells revealed that disruption of microtubules caused a severe delay in enucleation, but did not affect the final extent of enucleation. This indicates that microtubules are required for, but not essential to, enucleation. Currently it is unknown how microtubules regulate PI3K activation during enucleation. Interestingly, biochemical analyses showed that the regulatory subunits of PI3K can bind to tubulin (Inukai et al., 2000; Kapeller et al., 1995), but we do not know whether this occurs in late erythroblasts or whether this affects activation or localization of PI3K.

Interestingly, our results show that cell polarization is initiated through repositioning of the nucleus but not MTOC during enucleation, similar to that seen in migrating fibroblasts (Gomes et al., 2005; Schmoranzer et al., 2009). Nuclear repositioning occurs concomitantly with the rearrangement of the microtubule array that leads to asymmetric distribution of actomyosin and thus the generation of asymmetric force (see below). This force most probably drives the repositioning of the nucleus, somewhat similar to that seen in migrating cells where nuclear movement needs actin flow (Gomes et al., 2005).

In comparison with migrating cells, in late erythroblasts the nucleus moves only a short distance (~1 μm) for the cell to become polarized. Then why do late erythroblasts need a long time (2–3 hours) to establish polarization? This is probably because either rearrangement of the microtubule array or stable association of the nucleus with the plasma membrane, or both, might be a relatively slow event.

Our proposed model for erythroblast enucleation is summarized in Fig. 8. The reorientation from a radial microtubule array to an asymmetric array somehow results in asymmetric activation of PI3K. Asymmetric accumulation of the PI3K products in the plasma membrane then induce asymmetric actin assembly and disassembly by stimulating their effectors, such as α-actinin (Franke et al., 1997), profilin (Lassing and Lindberg, 1985) or gelsolin (Janmey et al., 1987), as well as Rac1 activity (Ji et al., 2008) probably through activation of its upstream regulator (Han et al., 1998; Kunisaki et al., 2006). This asymmetric actin assembly and disassembly and actomyosin contractions promote nuclear displacement as well as nuclear extrusion.

In cells defective in this microtubule-regulated PI3K-dependent polarization mechanism, enucleation could be achieved by random cytoplasmic contractions alone but in a highly ineffective and slow manner. It is likely that such defects in enucleation cause some proteins to be wrongly sorted to the reticulocytes or the extruded nucleus, leading to the formation of abnormal reticulocytes.

Our results suggest that enucleation is carried out in a way mechanistically similar to directed cell migration. Some migrating cells form a polarized bleb-like protrusion in order to move forward (Charras and Paluch, 2008; Haston and Shields, 1984; Paluch et al., 2006), and the difference between the role of the bleb-like protrusion in cell migration and enucleation is probably influenced by differences in their cytoskeletal organization. The unexpected findings in the present study thus begin to elucidate the regulatory and mechanistic mechanisms underlying mammalian erythroblast enucleation.

Pregnant C57 BL/6J mice were purchased from the Genome Institute of Singapore and the Center for Animal Resources (National University of Singapore). The purification and culturing of mouse fetal liver erythroblast precursors (CFU-e; TER119-negative cells) were performed as previously described (Ji et al., 2008; Zhang et al., 2003). For retroviral gene transfer experiments, we used the murine stem cell retroviral vector system (MSCV; Clontech, BD Biosciences). Infection of primary cells with mouse-specific retroviruses encoding fusion proteins of (3)Akt(PH)–GFP (Luo et al., 2005), (3)Akt(PH^R25C)–GFP, mCherry, RFP or GFP–γ-tubulin was performed as previously described (Ji et al., 2008). All animal experiments were performed according to the relevant regulatory standards.

For phase-contrast and bright-field time-lapse live-cell imaging, purified TER119-negative cells grown on fibronectin-coated glass chamber dishes (McKenna and Wang, 1989) were maintained at 37°C in a custom-made incubator built on top of an Axiovert 135 or Axiovert 200M inverted microscope (Carl Zeiss) and viewed with a 100×, Plan Apo NA 1.25 or 100×, NA 1.30, Plan-NEOFLUAR lens. Images were acquired with a video camera (Mintron) or a cooled charge-coupled device camera (CoolSNAP_HQ, Roper Scientific) and processed with MetaView (Universal Imaging) or Image-Pro Plus software. Images of cells expressing (3)Akt(PH)–GFP, cells expressing (3)Akt(PH)–GFP and mCherry or RFP, cells expressing (3)Akt(PH^R25C)–GFP and cells expressing GFP–γ-tubulin were acquired using an Axiovert 200M inverted microscope (Carl Zeiss) equipped with 100×, NA 1.4 Plan-Apochromat lens, a spinning disk confocal (Yokogawa CSU-21) scan head and Hamamatsu Orca-ER cooled CCD camera. Bright field images (Fig. 1B,D) were uniformly filtered to reduce noise using ImageJ (NIH, Bethesda, MD).

For imaging of F-actin, stacks of images acquired using a LSM 510 Meta confocal microscope system (100×, NA 1.4 Plan-Apochromat lens) were collected at a vertical interval of 0.75 μm and processed with IMARIS (Bitplane, Scientific Solutions).

For imaging of microtubules, Z-stacks of wide-field images were collected at intervals of 0.3 μm, were deconvoluted (for Fig. 3A) using AutoQuant (Media Cybernetics) and processed with a pattern recognition program that detects linear structures as described previously (Wang, 2003). For imaging of EB1 and pericentrin, stacks of wide-field images were collected at vertical intervals of 0.3 μm and projected with a maximum-intensity algorithm using ImageJ. For γ-tubulin, single wide-field images were collected.

Immunofluorescence staining was performed according to the methods previously described (Wheatley and Wang, 1998). We used primary antibodies against rabbit Ser19-phosphorylated myosin light chain II (1:50, Cell Signaling), mouse anti-β-tubulin (1:1000, AbCam), mouse anti-EB1 (1:50, BD Biosciences, Pharmingen), mouse anti-γ-tubulin (1:1000, Sigma); rabbit anti-pericentrin (1:700, Abcam). Secondary antibodies Alexa-Fluor-488-conjugated goat anti-rabbit IgG, Alexa-Fluor-488-conjugated goat anti-mouse IgG and Alexa-Fluor-546-conjugated goat anti-mouse IgG (Molecular Probes), all diluted 1:100. To observe F-actin, Rhodamine–phalloidin (Molecular Probes) was used at 1:300 dilution. Hoechst 33258 (1 μg/ml) was used to locate the nucleus.

Cytochalasin D (Sigma), blebbistatin (Toronto Research Chemicals), nocodazole (Sigma), LY294002 (Invitrogen) and wortmannin (Sigma) were dissolved in DMSO to make stock solutions of 25 mM, 100 mM, 3.3 mM, 50 mM and 1 mM, respectively, and applied to a culture dish at final concentration of 2 μM, 100 μM, 10 μM, 50 μM and 1 μM, respectively. Note that >10 μM nocodazole was required to completely depolymerize microtubules in late erythroblasts.

To measure the size of the nuclear area, the area of the ellipse corresponding to the shape of the nucleus was calculated using ImageJ. To track the distance between the centroids of the cell and the centroids of the nucleus or the MTOC, each perimeter was drawn and the centroid corresponding to the nucleus and the cell was calculated using ImageJ. The distance between these centroids was calculated.

We thank TLL Microscopy and Animal Care facilities and Shazmina Rafee for their technical assistance, Koichi Okumura (Cancer Science Institute of Singapore) for helpful suggestions, and G. Wright and J. Brocher for advice on image processing. We thank all members of the Murata-Hori laboratory for discussions and/or critical reading of the manuscript.