Signaling and cytoskeletal requirements in erythroblast enucleation

Erythropoiesis in mammals concludes with the dynamic process of enucleation, by which the orthochromatic erythroblast generates a reticulocyte that will mature to become a red blood cell, and a pyrenocyte, a membrane-encased nucleus surrounded by a thin rim of cytoplasm.^1-3  After enucleation, the reticulocytes are released into the bloodstream, whereas the pyrenocytes expose apoptotic signals on their surface, resulting in engulfment and degradation by the central macrophage of the erythroblastic island.^3,4  The mechanism of enucleation has been a long-standing matter of investigation and remains controversial.

Early electron microscopy studies suggested that enucleation may be analogous to cytokinesis, pointing to the resemblance of the cytoplasmic constriction between incipient reticulocyte and pyrenocyte to the cleavage furrow at the equatorial region of a mitotic cell.⁵  Koury et al demonstrated with electron and immunofluorescent microscopy, using mouse splenocytes infected with the anemia-inducing strain of Friend virus, that F-actin bundles concentrate at the furrow behind the extruding nucleus, and that cytochalasin-D, a potent inhibitor of actin polymerization, inhibits enucleation.⁶  In parallel, a quantitative study by Chasis et al showed that inhibition of microtubule polymerization by colchicine stalls enucleation in vivo and in vitro in rat bone marrow.⁷  A recent study by Keerthivasan et al using primary mouse and human erythroblasts supported the hypothesis that vesicle trafficking and subsequent vacuole coalescence provide additional membrane for the separating nucleus and reticulocyte.^8,9  Actin polymerization was again seen to play a role by Ji et al who found that Rac GTPase deregulation by overexpression of either dominant negative or constitutively active mutants of Rac1 or Rac2 or by pharmacologic inhibition by the Rac-specific inhibitor NSC23766^10,11  decreases enucleation of mouse fetal liver erythroblasts in culture. Moreover, down-regulation of mDia2, a downstream effector of Rho GTPases, by small interfering RNA (siRNA) blocks erythroblast enucleation.¹⁰ 

Rac GTPases are subfamily members of the Rho GTPases family and Ras superfamily. Rac1 and Rac2 GTPases have been shown to play distinct and overlapping roles in hematopoietic stem/progenitor and mature blood cells.^12-14  Rac1, Rac2, and Rac3 are known to regulate, among other processes, actin structures and vesicular transport pathways.¹⁵  As noted earlier in this section, both actin and vesicle trafficking have been implicated in separate proposed models of the enucleation mechanism. On the other hand, cytokinesis has been shown to require microtubules, actin, and vesicle trafficking along with associated signaling proteins in a sequential rather than mutually exclusive manner.^16,17  In the present study, we developed a novel analysis protocol of high-speed cell imaging in flow (using the Imagestream^X imaging-flow cytometer; Amnis) to identify a population of enucleating erythroblasts from mouse bone marrow or spleen and study the distribution of cytoskeletal components during enucleation. We used erythroblasts genetically deficient of all 3 mammalian Rac GTPases along with wild-type (WT) control erythroblasts, and WT erythroblasts treated with pharmacologic inhibitors of cytoskeletal or lipid raft assembly to determine the roles of these elements in the enucleation process. We demonstrate a distinct role of microtubules in the establishment of cell polarity before enucleation, with subsequent formation of a contractile actomyosin ring and coalescence of lipid rafts between incipient reticulocyte and pyrenocyte, controlled by Rac GTPases. These data reveal a coordinated assembly in time and space of microtubules, F-actin, and lipid rafts conducting the enucleation process.

Mx1Cre^Tg/+;Rac1^flox/flox;Rac2^−/− mice, bred in a 129Sv and C57BL/6J background, were generated as described previously (supplemental Methods, available on the Blood Web site; see the Supplemental Materials link at the top of the online article).^14,18  These mice were bred with Rac3^−/− mice¹⁹  to create Mx1Cre^Tg/+;Rac1^flox/flox;Rac2^−/−;Rac3^−/− mice. Cre-mediated recombination to induce hematopoietic specific deletion of Rac1 (resulting in Rac1^Δ/Δ;Rac2^−/− and Rac1^Δ/Δ;Rac2^−/−;Rac3^−/− mice, respectively) was carried out by polyinosinic-polycytidylic acid (pI-pC; Amersham-Pharmacia Biotech) treatment using 4 to 6 intraperitoneal injections of pI-pC.^14,20  All animal protocols were approved by the Institutional Animal Care and Use Committee of Cincinnati Children's Hospital Medical Center.

To study erythroid enucleation, we used 2 in vitro systems. The first was a modification of the long-term ex vivo erythroid differentiation culture protocol described by Giarratana et al,²¹  adapted for mouse cells. Low-density bone marrow cells were cultured in erythroblast growth medium (StemPro-34 with 2.6% StemPro-34 supplement; Invitrogen), 20% BIT 9500 (StemCell Technologies), 900 ng/mL ferrous sulfate, 90 ng/mL ferrous nitrate, 10⁻⁶M hydrocortisone, penicillin/streptomycin, l-glutamine), in 3 steps. In the first proliferative step (days 1-5), the medium was supplemented with 100 ng/mL SCF, 5 ng/mL IL-3, and 2 IU/mL human erythropoietin (Amgen). SCF and IL-3 retard differentiation of the erythroid progenitors while they promote proliferation,²²  facilitating expansion of erythroid precursors by adding all 3 cytokines fresh every other day. In the second step (days 6-7), the cells were resuspended at a concentration of 5 × 10⁵ cells/mL, in wells coated with fibronectin at 50 μg/mL, in fresh medium supplemented with only erythropoietin. The medium was changed on day 7 to fresh medium without cytokines, and the cells proceeded to enucleation (days 8-9). For Rac inhibition studies, 50 or 100μM of the Rac specific inhibitor NSC23766¹¹  was added on day 7 and replaced with fresh NSC23766 at the same concentration with medium change on day 8.

The second system is the fast enucleation assay described by Yoshida et al,³  with slight modifications. Briefly, stress erythropoiesis was induced by phlebotomy of 500 μL whole blood with equal volume of normal saline replacement intraperitoneally in adult WT C57BL/6 mice. After 4 days, the mice were killed to obtain the spleen, which was gently homogenized. Splenocytes were separated from parenchyma by filtration through a 40-μm cell-strainer (BD Biosciences), suspended in IMDM plus 2% FBS, separated by Ficoll gradient using Histopaque 1.083 g/mL (Sigma-Aldrich), and treated with red blood cell-lysis buffer (Sigma-Aldrich). The isolated low-density splenocytes, highly enriched in erythroblasts after the phlebotomy, were suspended in erythroblast growth medium supplemented with erythropoietin (10 IU/mL) and cultured on plastic, overnight at 37°C. The next day, cells were lifted using PBS plus 10mM EDTA on ice, washed, and plated at a concentration of 10⁶ cells/well in cytokine-free erythroblast growth medium, in a 24-well plate coated with MS5 cells and incubated to enucleation for 6 to 8 hours (time guided by microscopic observation of ∼ 30%-40% enucleation in the untreated WT sample) at 37°C, with and without the following pharmacologic inhibitors: cytochalasin-D, colchicine, filipin (Sigma-Aldrich), ML-7 (EMD4 Biosciences; Merck KGaA), taxol (Cayman Chemical), or NSC23766, at the concentrations indicated in the corresponding figures.

All experiments involving erythroblasts from Rac1^Δ/Δ;Rac2^−/− or Rac1^Δ/Δ;Rac2^−/−;Rac3^−/− mice (see Figures 1A-B and 2,Figure 3–4) were performed with the long-term enucleation assay to avoid phlebotomy in these mice within the first week after pI-pC induction, a procedure that would increase their already high mortality rate. This assay allowed production of sufficient number of enucleating cells for protein pull-down assays to determine activity of Rac GTPases. The remainder of the figures present experiments that were performed with the fast enucleation assay, which allowed evaluation of inhibitors, such as colchicine and ML-7, which appear to cause cell toxicity with more than 12-hour incubations. Cytospins were prepared at 400 rpm for 3 minutes in a Cytospin-4 Cytocentrifuge (Thermo Shandon), stained with Wright stain (Harleco EMD), and photographed using an Olympus BX51 microscope equipped with an Optronics Macrofire Camera and a PictureFrame Application Version 2.3 software (Optronics).

To study the enucleation process quantitatively, we evaluated the percentage of enucleated cells at the conclusion of each experiment by flow cytometry. The cells from each well were collected by washing with PBS plus 10mM EDTA on ice and stained with PE-Cy7-anti–Ter119 (BD Biosciences) in 100 μL of PBS plus 0.5% BSA (FACS buffer), for 20 minutes. After gentle washing to remove unbound antibody, the cells were resuspended in 10mM HEPES buffer, pH 7.4, containing 140mM NaCl and 5mM MgCl₂ with 0.25μM SYTO16 (Invitrogen) or FACS buffer with Hoechst 33342 (Invitrogen) for nuclei detection. In both cases, cells were stained with 7-amino-actinomycin D for exclusion of dead cells from the investigation. FACS analysis was conducted on a FACSCanto Flow Cytometer or an LSR-II Flow Cytometer, with FacsDiva Version 6.1.3 software (BD Biosciences).

Measurement of the active (GTP-bound) Rac proteins was performed as previously described^18,23  (detailed in supplemental Methods).

The distribution of cytoskeletal molecules within enucleating erythroblasts was studied by Imagestream^X (Amnis), which combines flow cytometry and microscopy (40×/numerical aperture 0.75 and 60×/numerical aperture 0.9 objective lens) capabilities. Erythroblasts were collected at the completion of fast enucleation assay using PBS plus 10mM EDTA on ice, were pelleted for 30 seconds at 2000g in a bench-top centrifuge, and fixed in PBS containing 4% formaldehyde for 20 minutes at room temperature. Formaldehyde was removed by centrifugation (30 seconds at 2000g) and aspiration of the supernatant, and the cell pellet was cooled on ice for 10 minutes, before permeabilization by consecutive suspensions in ice-cold (kept at −20°C) 50% acetone, then 100% acetone, and then again 50% acetone solution,²⁴  followed by one wash with FACS buffer. Cells were resuspended in FACS buffer and incubated with PE-Cy7– or PE-anti–Ter119 antibody (BD Biosciences) and either AlexaFluor-488–anti–β-tubulin antibody (BD Biosciences), AlexaFluor-488–conjugated or rhodamine-phalloidin (Invitrogen), AlexaFluor-594–conjugated cholera toxin subunit B (CTB; Invitrogen), anti–pMRLC (Ser19) primary antibody (Cell Signaling Technology), followed by AlexaFluor-488–conjugated secondary antibody (Invitrogen), and anti–γ-tubulin primary antibody (Cell Signaling Technology) followed by AlexaFluor-555–conjugated secondary antibody (Invitrogen). After washing once with FACS buffer, the nuclear stain Draq5 (Cell Signaling Technology) was added at a concentration of 40μΜ in FACS buffer and the samples were processed by Imagestream^X. Approximately 10 000 events per experiment were collected and analyzed with the associated Image Data Exploration and Analysis software (IDEAS; Amnis). Bright Detail Intensity R3 was determined in the CTB-stained enucleating erythroblasts by computing the intensity of localized bright spots within the image that are 3 pixels in radius or less after the local background around the spots is removed (similarly to the “top-hat morphologic transformation” in microscopy image analysis using a 3-pixel radius structuring element and then computing the intensity of the resulting image).

Immunofluorescence staining and confocal microscopy was as previously described.^18,25  (detailed in supplemental Methods)

Statistical significance for enucleation efficiency differences was evaluated using Student t test for unpaired data with equal variance (results presented as mean ± SEM). The Imagestream^X data of the individual values of centroid distances and bright detail intensities in the experiments shown in Figures 5 and 6 were exported for statistical evaluation. Both delta centroid distance and bright detail intensity measurements take non-negative values, and their probability distribution is asymmetric with a skew toward large values, indicating non-normality. Therefore, to assess the statistical significance of the difference in the medians of the test and control groups, we used the Wilcoxon rank-sum test (nonparametric test for 2 independent samples), which makes no assumptions about the underlying probability distributions and is thus appropriate for non-normal data. For each set of measurements, we report the mean and the median for each group and the 2-tailed P value according to the Wilcoxon rank-sum test.

To study erythroid enucleation quantitatively, we used 2 in vitro systems detailed in “In vitro erythropoiesis and enucleation,” a long-term ex vivo erythroid differentiation culture protocol,²¹  adapted for mouse cells, and a fast enucleation assay as described by Yoshida et al,³  with slight modifications. We evaluated the percentage of enucleated cells at the conclusion of each experiment by flow cytometry. After gating on Ter119-positive cells, the SYTO16-low and negative population represents the enucleated erythrocytes (Figure 1A-B).

The low frequency of erythroblasts engaged in enucleation that can be studied microscopically is often a limiting factor in enucleation studies. Consequently, we sought to develop a method that would enable us to identify and study a significant number of enucleating erythroblasts under specific conditions. To this end, we expanded a method previously described by McGrath et al using the Imagestream^X imaging flow cytometer (Amnis) that allows separation of basophilic, polychromatophilic, and orthochromatic erythroblasts by gating the cells according to their size and intensity of Ter119 staining (Figure 1C).²⁶  We used the aspect ratio and δ centroid XY, morphologic cell parameters measurable through the use of the associated software IDEAS (Amnis). The aspect ratio is the ratio of the minor cell axis divided by the major axis and describes how round or oblong the cell is. The δ centroid XY parameter measures the distance between the centers of 2 chosen (X and Y) stains, in this case, Ter119 and Draq5. The population with low aspect ratio and high δ centroid Ter119-Draq5 value is highly enriched in enucleating cells, in which the nucleus is moving away from the main body of an elongated cell (Figure 1D). This method facilitated the visualization of fixed enucleating cells (Figure 1Eiv-vi) and consequently the study of localization of cytoskeleton and signaling components during enucleation. In the representative experiment shown in Figure 1C through E, we gated 103 enucleating cells within the population encircled by the red line. Thirty-one of these 103 events were at a stage during enucleation and at an imaging section where F-actin was visible as a ring in the cleavage furrow between incipient reticulocyte and pyrenocyte, suggesting that visualization of a significant number of enucleating events is necessary to ascertain the distribution of structures during this highly dynamic process (supplemental Figure 1).

Deregulation of Rac GTPases with overexpression of either dominant negative or constitutively active mutants of Rac1 or Rac2 or by pharmacologic inhibition by the Rac-specific inhibitor NSC23766¹¹  has been shown by Ji et al to decrease enucleation of mouse fetal liver erythroblasts in culture.¹⁰  However, we found that the in vitro enucleation efficiency of Rac1^Δ/Δ;Rac2^−/− erythroblasts, produced in Mx1Cre^Tg/+;Rac1^flox/flox;Rac2^−/− adult mouse bone marrow after inducing deletion of the floxed Rac1 gene in vivo by pI-pC injections, was increased compared with the enucleation of WT erythroblasts in parallel long-term erythropoiesis cultures as quantitated by flow cytometry (Figure 2A-B). Because we and others have demonstrated compensatory cross-talk between members of the Rho GTPase family (Rho, Rac, and Cdc42),^28-30  we evaluated the protein levels of these GTPases in Rac1^Δ/Δ;Rac2^−/− versus WT erythrocytes by immunoblotting. Whereas RhoA and Cdc42 was unchanged (data not shown), Rac3 was significantly increased in Rac1^Δ/Δ;Rac2^−/− erythrocytes (Figure 2C-D). Thus, it appears that Rac1 and Rac2 ablation is not sufficient to impair enucleation because of compensatory up-regulation of Rac3 expression and/or protein stabilization. To determine the physiologic relevance of increased Rac3 level in Rac1^Δ/Δ;Rac2^−/− erythroblasts, we used the Rac-specific inhibitor NSC23766¹¹  to simultaneously inhibit Rac1, Rac2, and Rac3. In agreement with the results of Ji et al in fetal liver cells,¹⁰  NSC23766 inhibited enucleation of adult bone marrow erythroblasts by more than 60% in WT and by more than 45% in Rac1^Δ/Δ;Rac2^−/− samples, in a dose-dependent fashion (Figure 2E). The inhibition of enucleation by NSC23766 was accompanied by parallel inhibition of Rac1, Rac2, and Rac3 activities, as confirmed by pull-down assays and consequent immunoblotting (Figure 2F-G). The enucleation efficiency of Rac3^−/− erythroblasts was evaluated by flow cytometry in fast enucleation assays in vitro and was found similar to that of WT erythroblasts (41.5% ± 2.5% vs. 43.7% ± 1.8%, n = 3, P = .14). These results indicate that the overall Rac GTPase activity plays a key role in regulating enucleation.

Rac GTPases are well-known regulators of actin in cells.¹⁵  We used the Imagestream^X imaging flow cytometer to study actin and myosin distribution in WT or Rac-deficient (after genetic or pharmacologic manipulation) enucleating erythroblasts produced in vitro in long-term enucleation assays. Actin filaments appeared to form a CAR between the nascent reticulocyte and the nucleus in WT erythroblasts, as described before,⁶  distributed in a similar pattern with actin in the cleavage furrow of a cell in cytokinesis (Figure 3A). In cytokinesis, furrow ingression is driven by assembly and contraction of actomyosin filaments.^16,31  Therefore, we examined whether myosin is also localized at the same area during enucleation. Labeling for phosphorylated myosin regulatory light chain (pMRLC) demonstrated that pMRLC and, consequently, activated myosin is contributing to the formation of CAR between reticulocyte and nucleus in WT enucleating erythroblasts, providing the necessary collaborator for actin to contract (Figure 3B). Parallel imaging of enucleating Rac1^Δ/Δ;Rac2^−/− erythroblasts demonstrated more intense membrane staining for actin in the incipient reticulocytes, as expected from the phenotype of Rac1^Δ/Δ;Rac2^−/− erythrocytes,¹⁸  as well as decreased intensity of staining for pMRLC. However, both actin and pMRLC were still able to assemble in CAR formation between reticulocyte and nucleus. In contrast, when NSC23766 was used to inhibit all Rac GTPases, neither F-actin nor pMRLC concentrated at the cleavage furrow to assemble a functional contractile ring; instead, these proteins continued to surround the nucleus (Figure 3A-B). This mislocalization probably results in the inhibition of enucleation that we observed by flow cytometric studies (Figure 2). Confocal microscopy of WT enucleating erythroblasts incubated without and with NSC23766 confirmed the persistent distribution of F-actin around the nucleus, when all Rac-GTPases are inhibited (Figure 3C). Staining of WT erythroblasts with both AlexaFluor-488–conjugated-pMRLC antibody and rhodamine-phalloidin for multiparameter high-speed cell imaging in flow showed colocalization of actin and myosin at the cleavage furrow between incipient reticulocyte and pyrenocyte (Figure 3D).

We investigated the significance of the contractile ring components in enucleation through flow cytometric analysis after incubation with pharmacologic inhibitors (Figure 3E). Inhibition of actin filament polymerization by cytochalasin-D at doses of 5, 10, and 20μM resulted in a statistically significant and dose-dependent decrease in enucleation efficiency, in agreement with data by Koury et al,⁶  leveling out by 20μM. Inhibition of MRLC phosphorylation through the use of the myosin light chain kinase (MLCK) inhibitor ML7 had a strong negative impact on the number of erythroblasts that achieved enucleation. Use of both cytochalasin-D and ML7 did not produce additive or synergistic results, indicating that both inhibitors target the common step of CAR formation. Inhibition of all Rac GTPases with NSC23766 led to decreased enucleation efficiency, interestingly to a greater degree than cytochalasin-D alone, suggesting the possibility that Rac GTPases affect enucleation not only via actin regulation. None of the inhibitor concentrations used in these studies led to cell apoptosis, as measured by 7-amino-actinomycin D staining, or to any morphologic alteration of the cells as evaluated in cytospin preparations, and their inhibition of enucleation was reversible on their removal from the medium (supplemental Figure 2A). We further confirmed the requirement of Rac GTPases for appropriate actin distribution during enucleation and ruled out off-target effects of NSC23766 by studying the enucleation efficiency of Rac1^Δ/Δ;Rac2^−/−;Rac3^−/− erythroblasts in long-term erythropoiesis cultures in vitro (Figure 4A) and evaluating the actin distribution of these cells by Imagestream^X (Figure 4B). The abnormal F-actin distribution surrounding the nucleus in those cells was the same as in the NSC23766-treated cells. Hence, Rac GTPases regulate enucleation, at least in part, through the organization of a contractile actomyosin ring.

Inhibition of microtubule polymerization in rat bone marrow cells in vivo or in vitro by colchicine has been shown by Chasis et al to cause an increased number of erythroblasts arrested at the act of nuclear extrusion.⁷  We also observed inhibition of enucleation by colchicine and taxol in the fast enucleation assay in vitro (Figure 5A), but there was no difference in staining with AlexaFluor-488–anti–β-tubulin within the enucleating erythroblasts exposed to 5μM colchicine for 6 hours versus the control sample (data not shown). However, when we specifically analyzed the population of orthochromatic erythroblasts, we could see polarized microtubule formation visible in control WT erythroblasts, whereas β-tubulin was diffusely stained in WT orthochromatic erythroblasts incubated with colchicine (5μM) for 6 hours in vitro as seen by Imagestream^x (Figure 5B). This unipolar microtubule assembly embracing the nucleus was also visualized in WT erythroblasts by confocal microscopy (Figure 5C). Incubation with colchicine that inhibits microtubule polymerization caused a diffuse staining for β-tubulin, whereas incubation with taxol that inhibits microtubule depolymerization produced thickened microtubule bundles. Microtubules radiated from a γ-tubulin–rich area within the narrow area between nucleus and plasma membrane as seen by Imagestream^x, whereas the nucleus moved toward the opposite direction for establishment of erythroblast polarity (Figure 5D). Statistical evaluation of the Imagestream^X data, by measuring the parameter delta centroid BF-DRQ5 between the center of the cell body as seen in bright-field and the center of the nuclear staining achieved with Draq5, demonstrated that inhibition of tubulin polymerization by colchicine abolished cell polarization with the nucleus remaining at a central location in the cell (Figure 5E). Moreover, when adding colchicine in a fast enucleation assay after pretreatment with cytochalasin-D and/or ML7 to first inhibit the contractile ring, no additive inhibitory effect on enucleation was observed (supplemental Figure 2B). These results indicate the critical role of microtubules in enucleation at a stage preceding contractile actomyosin ring formation.

Cholesterol-rich liquid-ordered plasma membrane microdomains (lipid rafts) are well known as platforms of signaling molecules and are essential for completion of cytokinesis after the formation of the actomyosin ring in the cleavage furrow.¹⁶  Rac GTPases have been shown to associate with lipid rafts when they bind to membrane to initiate signaling.^32,33  Because enucleation is regulated by Rac GTPases and uses an actomyosin ring in a cytokinesis-like mechanism, we investigated the potential role of lipid rafts during enucleation. Staining of erythroblasts incubated in the fast enucleation assay with fluorescent CTB, which binds to the lipid raft marker G_M1-ganglioside, revealed that lipid rafts do merge in the furrow between incipient reticulocyte and pyrenocyte (Figure 6A). Inhibition of Rac GTPases with NSC23766, apart from inhibition of enucleation and altered distribution of actomyosin, led also to the absence of lipid rafts between reticulocyte and pyrenocyte, within the elongated cells attempting enucleation (Figure 6A). The clustering of lipid rafts, as seen by evaluating bright detail intensity, decreased in enucleating erythroblasts incubated with NSC23766 compared with the control erythroblasts (Figure 6B). Filipin, an inhibitor of lipid rafts through cholesterol depletion, has been shown to inhibit cytokinesis.¹⁶  Incubation of WT erythroblasts in vitro with filipin, in the fast enucleation assay, inhibited enucleation in a dose-dependent fashion. Moreover, combining cytochalasin-D and filipin had a synergistic effect, reaching a similar level of inhibition with the Rac-inhibitor NSC23766, implicating that Rac GTPases probably control both F-actin and lipid raft assembly during enucleation.

Erythroblast enucleation is a dynamic, fast-moving process that lasts approximately 10 minutes.⁴  Because limited synchronization can be attained in either in vitro erythropoiesis cultures or by stress erythropoiesis in vivo, it has been challenging to evaluate morphologically a significant number of enucleating erythroblasts. Using Imagestream^X (Amnis), we developed a novel method to identify a population enriched in enucleating erythroblasts, based on quantifiable morphologic cell parameters. By focusing on the elongated cells with increased distance between the centers of Ter119⁺ staining and Draq5⁺ staining, we were able to examine the distribution of cytoskeleton and signaling components in fixed enucleating erythroblasts, collecting images of at least 80 events of interest in each standardized condition studied. From these, we had at least 30 images per experiment, which were at the appropriate enucleation stage and the appropriate focus for the cytoskeleton component in question to be evaluable.

By this method, we were able to visualize clearly F-actin structures concentrated in a ring (CAR) between incipient reticulocyte and pyrenocyte, in agreement with the images previously shown by Koury et al.⁶  Moreover, we demonstrated that pMRLC colocalized with actin at the same area, indicating that a functional actomyosin ring is being formed during enucleation. Nonmuscle myosin IIB was recently shown to participate in enucleation.³⁴  The exact mechanism of actin-myosin contraction, mediated either by myosin's head,^17,35,36  or by a tail-hinge apparatus,³⁷  but in any way regulated by MRLC phosphorylated at Ser19, a target of Rho-kinase (ROCK) or MLCK,³⁸  remains to be elucidated. Using our fast in vitro enucleation assays, we evaluated the action of cytochalasin-D and ML7, inhibitors of actin polymerization and MLCK, respectively. Both inhibited enucleation in a dose-dependent and reversible manner without a synergistic effect when combined, indicating that they both inhibit the same process during enucleation. The inhibition we observed with these inhibitors was not complete, especially for cytochalasin-D,^6,8  probably because of the high enucleation efficiency (30%-70%) in our assays, attained by using stroma cells. A more complete inhibition is easier to achieve when control samples have enucleation efficiency of 10% to 18% as previously described^6,8 ; however, we observed that experiments terminated at a low enucleation rate gave highly variable results.

We observed that inhibition of Rac GTPases by the pan-Rac inhibitor NSC23766 suppressed enucleation, consistent with previous report by Ji et al,¹⁰  and in agreement with our earlier findings that Rac GTPases control F-actin polymerization in erythrocytes.¹⁸  Multiparameter high-speed cell imaging in flow revealed that genetic or pharmacologic inhibition of all Rac GTPases was associated with failure to assemble F-actin and pMRLC into creating CAR next to the nucleus that had moved appropriately within the elongated cell (Figures 3A and 4B). Hematopoietic cells with genetic deletion of Rac1 and Rac2 GTPases were able to differentiate in vitro to erythroblasts, enucleating at least as efficiently as WT erythroblasts, because of concurrent Rac3 up-regulation. It is intriguing that Rac1^Δ/Δ;Rac2^−/− erythroblasts exhibited levels of enucleation higher than WT erythroblasts in erythropoiesis cultures. This finding implicates that the pathways regulating enucleation involve multiple signaling and cytoskeletal components. Deletion of some of these components at the genetic level may be associated with up-regulation of other genes at a parallel or downstream level that leads to overcompensation. Such an overcompensation in quantity may happen at the expense of quality, as it is suggested by the small, fragile reticulocytes produced by Rac1^Δ/Δ;Rac2^−/− erythroblasts.¹⁸  Erythroblast enucleation is a process necessary for mammalian survival, as indicated by the embryonal lethality of mouse mutants with genetic deletion of erythroblast-macrophage protein or of DNase II, where a significant number of fetal erythroblasts fail to enucleate.^39,40  Enucleation in vitro has a significantly decreased efficiency compared with the in vivo process. This superb in vivo efficiency is probably the result of interactions among the erythroblasts and of erythroblasts with the central macrophage and extracellular matrix within the erythroblastic islands.¹  Rac GTPases are activated by cell-receptor-mediated signaling,³²  and their optimal regulation of enucleation may depend on cell-cell and cell-extracellular matrix interaction within the erythroblastic island. Further studies will be required to evaluate the multiple processes collaborating to produce efficient enucleation.

We did observe inhibition of enucleation after incubation of erythroblasts with colchicine, in agreement to the study by Chasis et al,⁷  but in contrast to 2 other previous studies.^6,8  Optimization of our fast enucleation assay by adding stroma cells to achieve enucleation efficiency of higher than 30% in the untreated sample allowed us to observe the inhibitory effect of colchicine consistently. Although we cannot completely exclude the possibility that microtubule inhibition decreased enucleation by effects to the stromal cells, we were able to see altered microtubule distribution in the erythroblasts after incubation with colchicine or taxol (Figure 5B-C). Organized microtubules, labeled with AlexaFluor-488–anti–β-tubulin, were visible in the population of orthochromatic erythroblasts in the untreated sample, inducing cell polarity in preparation for enucleation. In contrast, β-tubulin was diffusely distributed in the cytoplasm in colchicine-treated samples. The distance between the center of the cell as seen in bright-field and the center of the nucleus stained by Draq5, as measured by Imagestream^X, was decreased in colchicine-treated versus control samples, indicating that inhibition of microtubules results in a significant decrease in erythroblast polarity and inhibition of enucleation in the subsequent step.

The essential role of Rac GTPases in enucleation is exerted at least in part via their regulatory role on actin polymerization. However, we noted that NSC23766 caused a more efficient inhibition of enucleation than the maximum inhibitory effect we could attain with cytochalasin-D (Figure 3E). Active Rac is known to bind to low-density, cholesterol-rich membranes identified by the lipid raft marker G_M1-ganglioside.^32,33  Such lipid domains have been shown to cluster in the cleavage furrow of dividing cells.¹⁶  Accordingly, we found that lipid rafts did merge in the furrow between incipient reticulocyte and pyrenocyte, as visualized by Imagestream^X after labeling of G_M1-gangliosides with fluorescent CTB. Inhibition of Rac GTPases decreased clustering of lipid rafts. Moreover, inhibition of lipid raft organization by filipin, through cholesterol depletion, inhibited enucleation. The combination of cytochalasin-D with filipin had a synergistic effect in inhibiting enucleation, equivalent to the inhibition attained by Rac inhibition, suggesting that Rac GTPases may regulate enucleation through a combined effect on F-actin and lipid raft assembly. Lipid rafts are important for membrane trafficking,^41,42  and it has been proposed that they may target secretory vesicles to the division site at the time of membrane addition during cytokinesis.⁴³  Further work is needed to explore the potential association of these signaling cholesterol-rich lipid platforms with the clathrin-dependent endocytic vesicles identified to contribute to enucleation by Keerthivasan et al.⁸ 

Our data support the concept that erythroblast enucleation is a complex, multistep process that involves numerous cytoskeleton and signaling components and has several similarities with cytokinesis. We propose a coordinated interplay of microtubules at an early stage, with actin and lipid rafts under the regulation of Rac GTPases at a later stage (Figure 7). F-actin interacts with myosin to provide a contractile ring, after phosphorylation of MRLC, probably driven by different signaling pathways. Enucleation is a robust process, necessary for survival, optimal circulation, and oxygen transfer in mammals; therefore, some redundancy may have been developed in the mechanisms involved in vivo, as demonstrated by the up-regulation of Rac3 in Rac1^Δ/Δ;Rac2^−/− erythroblasts. Further research will be required to identify the detailed molecular events to gain a complete understanding of the enucleation process in erythropoiesis.

The authors thank Dr Adrienne Cox (University of North Carolina, Chapel Hill, NC) for the generous gift of Rac3-specific antibody, Dr Ivan de Curtis (San Raffaele Scientific Institute, Milano, Italy) for providing Rac3^−/− mice, the Flow Cytometry core of Cincinnati Children's Hospital Medical Center, and Richard Demarco, Sherree Friend, and Scott Mordecai from Amnis Corporation for expert technical support.

This work was supported by the National Institutes of Health (grants K08 HL088126, T.A.K.; R01 CA141341, Y.Z.; and 5R01DK062757-11, D.A.W.).

National Institutes of Health

Contribution: T.A.K., D.G.K., and S.P. designed and performed research and analyzed data; J.F.J. and C.E.H. performed research and analyzed data; S.M. performed the statistical analysis; J.A.C., D.A.W., and Y.Z. contributed instrumental suggestions and mentorship on the research design and analysis of data; and all authors wrote the manuscript.

Conflict-of-interest disclosure: The authors declare no competing financial interests.

Correspondence: Theodosia A. Kalfa, Division of Hematology/Oncology, Cincinnati Children's Hospital Medical Center, 3333 Burnet Ave, MLC 7015, Cincinnati, OH 45229-3039; e-mail: theodosia.kalfa@cchmc.org.