Formation of mammalian erythrocytes:chromatin condensation andenucleationPeng Ji1, Maki Murata-Hori2,3 and Harvey F. Lodish4,5,6

1 Department of Pathology, Feinberg School of Medicine, Northwestern University, Chicago, Illinois 60611, USA2 Temasek Life Sciences Laboratory, 1 Research Link, National University of Singapore, 117604, Singapore3 Department of Biological Sciences, National University of Singapore, Singapore, 117604, Singapore4 Whitehead Institute for Biomedical Research, Cambridge, Massachusetts 02142, USA5 Department of Biology, Massachusetts Institute of Technology, Cambridge, Massachusetts 02142, USA6 Department of Biological Engineering, Massachusetts Institute of Technology, Cambridge, Massachusetts 02142, USA

Review

In all vertebrates, the cell nucleus becomes highly con-densed and transcriptionally inactive during the finalstages of red cell biogenesis. Enucleation, the processby which the nucleus is extruded by budding off from theerythroblast, is unique to mammals. Enucleation hascritical physiological and evolutionary significance inthat it allows an elevation of hemoglobin levels in theblood and also gives red cells their flexible biconcaveshape. Recent experiments reveal that enucleationinvolves multiple molecular and cellular pathways thatinclude histone deacetylation, actin polymerization, cy-tokinesis, cell–matrix interactions, specific microRNAsand vesicle trafficking; many evolutionarily conservedproteins and genes have been recruited to participate inthis uniquely mammalian process. In this review, wediscuss recent advances in mammalian erythroblastchromatin condensation and enucleation, and concludewith our perspectives on future studies.

Mammalian erythroblast enucleationRed blood cells are continuously replenished; in humanstheir half-life is around 120 days. Erythropoietin (Epo), acytokine produced by the kidney in response to low oxygenpressure, is the principal regulator of red blood cell produc-tion. Epo binds to cognate receptors on Colony Forming UnitErythroid (CFU-E) progenitor cells and activates the JAK2protein tyrosine kinase and several downstream signaltransduction pathways. These prevent CFU-E apoptosisand stimulate its terminal proliferation and differentiation;during the ensuing three days, each CFU-E produces 30–50enucleated reticulocytes that are released into the circula-tion. The first phase of CFU-E erythroid differentiation ishighly Epo dependent, whereas later stages are no longerdependent on Epo but are enhanced by adhesion of erythro-blasts to a fibronectin substratum [1]. Consistent with this,Epo receptors are lost as erythroid progenitors undergoterminal proliferation and differentiation [2].

As judged by high throughput mRNA sequencing, theexpression of about 600 genes increases greater than

Corresponding authors: Ji, P. (peng-ji@fsm.northwestern.edu);Lodish, H.F. (lodish@wi.mit.edu).

0962-8924/$ – see front matter � 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.tcb.2011.0

twofold during terminal erythroid differentiation (P. Wonget al., unpublished data). These include many erythroid-important genes including a and b globins, heme biosyn-thetic enzymes, erythroid membrane and cytoskeletalproteins, and erythroid-important transcription factors.Concomitantly, the volume of the nucleus graduallydecreases, ultimately becoming approximately a tenth ofits original volume with highly condensed chromatin (Fig-ure 1). Although in all vertebrates the erythrocyte nucleusbecomes highly condensed, only mammalian erythroblastsenucleate (Box 1). Red cells in many non-mammalianvertebrates contain a special linker histone H5 that isnot seen in mammals [3–5].

Our understanding of mammalian erythroid enucle-ation has increased significantly since the results of mor-phological studies carried out decades ago were obtained[6–11]. Time-lapse live-cell imaging, for example, showsthat the erythroblast quickly extrudes its nucleus, sur-rounded by and closely apposed to the plasma membranethrough a bleb-like structure from a limited area of thecortex adjacent to the nucleus (Figure 2a). Notably, thenucleus becomes largely deformed during nuclear extru-sion and the outer nuclear membrane becomes closelyapposed to the plasma membrane (Figure 2b), which sug-gests a possible requirement for remodeling of the nuclearenvelop prior to enucleation. The process is completed by atype of cytokinesis that requires actin and other cytoskel-etal proteins, which are discussed later. Here, we discussthe many recent advances in understanding mammalianerythroblast chromatin condensation and enucleation, andconclude with a discussion of future studies that can clarifyour understanding of these processes.

Chromatin condensation and apoptosisDuring mammalian erythropoiesis, the chromatin grad-ually undergoes condensation; nuclear and chromatincondensation are thought to be critical for enucleation.Chromatin condensation occurs during other cellularprocesses such as apoptosis, and apoptotic mechanismsare known to play important roles in erythropoiesis(reviewed in [12]). Apoptotic mechanisms have been

4.003 Trends in Cell Biology, July 2011, Vol. 21, No. 7 409

Box 1. Vertebrate erythrocytes and evolution of

mammalian erythroblast enucleation

Erythroid cells in all vertebrates undergo gradual chromatin

condensation during erythropoiesis. The mature circulating red

blood cells vary in size across the vertebrates, ranging from over

50 mm in diameter in certain species of amphibians to less than

10 mm in mammals [73]. The size of the circulating red blood cells

correlates with the diameter of capillaries, as well as the capacity of

the vertebrate heart to pump blood to the extremities. In mammals,

the presence of four heart ventricles enables sufficient blood flow

through the microcirculation, where many capillaries possess a

diameter less than that of mature red blood cells. Therefore,

circulating mammalian red blood cells must change or deform their

shape to migrate through these small capillaries. Enucleation is

hypothesized to have been selected for during mammalian evolu-

tion to enhance blood cell circulation and prevent possible blockage

of small capillaries by deformed red cells. The lack of a nucleus is

also thought to provide more intracellular space for hemoglobin.

However, blood hemoglobin content and mean cell hemoglobin

concentration are similar between mammals and birds [73], which

could be as a result of the fact that in birds hemoglobin is also

localized in the nuclear space [73].

Review Trends in Cell Biology July 2011, Vol. 21, No. 7

implicated in the loss of the nucleus in mammalian lensepithelia and keratinocytes [13,14], although these pro-cesses are morphologically dissimilar to erythroid cellenucleation.

Proerythroblast Early erythroblast

Nucleus

Late e

Chromatin condensation

HDACs Gcn5

Myc

miR-191

Mxi1

Rio

Nu

Figure 1. Enucleation of mammalian erythroblasts requires multiple molecular and c

gradually condenses while the hemoglobin concentration gradually increases. Chromati

histone deacetylases (HDACs) and down-regulation of histone acetyltransferases (HA

decreases during erythropoiesis. In addition, miR-191, an erythroid specific microRNA

chromatin condensation could provide unknown signals to activate the Rac GTPases–mD

410

shRNA knockdown of caspase 3 in human erythroidcells leads to a significant reduction in enucleated cellswith no change in the fraction of apoptotic cells [15].However, this phenotype could be a consequence of theblockage of erythroid differentiation from pronormoblaststo basophilic normoblasts after down-regulation of caspase3 [15,16]. In addition, treatment of mouse spleen erythro-blasts with caspase inhibitors fails to block enucleation[17]. Furthermore, there is no evidence for caspase-inducedcleavage of target proteins during the late stages of eryth-ropoiesis [18]. These results indicate that erythroid chro-matin condensation and enucleation might not requirecaspases or other apoptotic proteins.

Modifications of the flexible ‘tails’ of several histones doplay important roles in the structure and stability ofchromatin. For example, histone acetylation by histoneacetyl transferases (HATs) tends to destabilize chromatin,which makes the DNA more accessible to transcriptionfactors. By contrast, histone deacetylation by histone dea-cetylases (HDACs) stabilizes chromatin and is critical forheterochromatin formation. A recent study demonstratedthat global levels of several acetylated histones drop dur-ing the differentiation of cultured mouse fetal liver ery-throblasts, especially H3K9Ac, H4K5Ac, H4K12Ac andH4K8Ac [19]. Furthermore, the level of Gcn5 protein, a

Hemoglobin

TRENDS in Cell Biology

Reticulocyte RBC

rythroblast

Rac

k3

cleus polarization

Contractile actin ring

mDia2

ellular pathways. During mammalian erythroblast differentiation, the chromatin

n condensation involves histone deacetylation that is controlled by up-regulation of

Ts), such as Gcn5. Gcn5 is directly regulated by c-myc, whose level gradually

, targets Mxi1 and Riok3 to regulate c-myc and Gcn5, respectively. Subsequently,

ia2 pathway, which is required for contractile actin ring formation and enucleation.

0 0.5 1

1.5 2 3.5

(min)(a)

(b)

2 μm 2 μm

TRENDS in Cell Biology

Figure 2. During enucleation, the nucleus is rapidly squeezed out from the cell,

forming a bleb-like structure. (a) Time-lapse live-cell imaging of late erythroblasts

undergoing nuclear extrusion. The nucleus is quickly squeezed out from a limited

area of the cortex facing the nucleus, forming the bleb-like structure (arrows). Note

that the nucleus becomes largely deformed during enucleation, and that a

cytokinetic-like furrow is observed at the neck region of the bleb-like protrusion

(arrowheads). Scale bar, 5 mm. (b) Electron microscope images of late erythroblasts

undergoing enucleation in an in vitro culture system. Late erythroblasts at early

(left) and late (right) stages of nuclear extrusion are shown. These images show

deformed nuclei and cytokinesis-like furrows similar to what has been seen in

erythroid cells from mice [10], which indicates that late erythroblasts in the in vitro

culture system reproduce the in vivo situation. The images were taken by T. Ramirez

and M. Murata-Hori. Scale bar, 2 mm.

Review Trends in Cell Biology July 2011, Vol. 21, No. 7

HAT, gradually decreases during erythropoiesis. Ectopicexpression of Gcn5, which makes chromatin more open,partially blocks chromatin condensation and subsequentenucleation [19]. In addition, Gcn5 is transcriptionallyactivated by c-Myc, whose level also gradually decreasesduring erythropoiesis. Expression of c-Myc at physiologicallevels during the terminal stages of erythropoiesis alsoblocks chromatin condensation and enucleation [19].

Consistent with this result, HDAC2, whose level grad-ually increases during erythropoiesis, is required for chro-matin condensation and enucleation; pharmacologicalinhibition of HDAC2 activity or shRNA knockdown ofHDAC2 blocks chromatin condensation and enucleationof cultured mouse fetal erythroblasts [20]. Similarly, treat-ment of Friend virus-infected murine spleen erythroblastswith HDAC inhibitors blocks enucleation [21]. These stud-ies establish the critical role of inhibiting histone acetyla-tion and activating histone deacetylation in terminalerythroid cell chromatin condensation and enucleation.We do not know whether the histones that are deacetylatedare localized in specific gene or chromosome regions or aredispersed throughout the genome.

Membrane and cytoskeletal proteins that mediateenucleation and asymmetrical cell divisionEarly studies indicated that enucleation involves a processof cytokinesis involving several membrane and cytoskele-ton proteins [9,10,22,23]. Mature red cells possess a uniquesubmembrane cytoskeleton that maintains the stabilityand flexibility characteristic of these cells. Essentially,

the heterodimeric a- and b-spectrins self-associate intoa2b2 tetramers, which are connected by actins to formpentagonal or hexagonal lattices [24]. Ankyrin–protein4.2 and protein 4.1–adducin complexes further mediatethe interaction of the spectrin–actin network with plasmamembrane proteins Band 3 (also called Anion Exchanger 1)and glycophorin, respectively. Altogether, these membraneand cytoskeletal proteins ensure the integrity of the ma-ture erythrocyte membrane (reviewed in [25]).

Functional disruptions of these proteins in humans canlead to hereditary spherocytosis, hereditary elliptocytosisor hereditary pyropoikilocytosis [26]. In light of the signifi-cance of these proteins, recent work showed that thenumber of normoblasts arrested at the stage of nuclearextrusion dramatically increases when a rat is treated withcolchicine, a microtubule-disrupting substance [23]. Inaddition, the F-actin inhibitor cytochalasin D causes acomplete inhibition of enucleation that could be reversedby washing out the cytochalasin D [22]. Furthermore,studies using yolk sac-derived erythroblasts showed thatactin and myosin become accumulated at the area betweenthe extruding nucleus and incipient reticulocyte, the so-called cortical actin ring [27].

One initial step in the enucleation process is that lateerythroblasts establish a polarization such that the nucle-us is displaced to one side of the cytoplasm [9,10] (Figure 2).Little is known about how this asymmetry is established.This process might involve extensive cytoskeletal andmembrane changes, including changes in the microtubulenetworks, and appears to share many features with thoseseen in migrating cells (J. Wang et al., unpublished data).

Rac GTPases play essential roles in enucleation throughthe activation of a downstream target, the formin mDia2,and subsequent formation of the contractile actin ring(Figure 1). Deregulation of Rac GTPases during the latestages of erythropoiesis completely blocked enucleation ofcultured mouse fetal erythroblasts without affecting nor-mal proliferation and differentiation [28]. The contractileactin ring formed on the plasma membrane of the late-stage erythroblasts at the boundary between the cyto-plasm and the nucleus of enucleating cells was disruptedwhen Rac GTPases were inhibited in the late stages oferythropoiesis. The activities of Rac GTPases are mediatedby the downstream target protein, mDia2, a formin proteinrequired for nucleation of unbranched actin filaments.These results reveal important roles for Rac GTPasesand mDia2 in enucleation of mammalian erythroblasts[28] (Figure 1).

Many other membrane and cytoskeletal proteins under-go reorganization during terminal erythropoiesis. Prior toand during enucleation, spectrin, actin, tubulin, ankyrin,glycophorin A and protein 4.1 all partition to the incipientreticulocyte, whereas no major erythroid membrane orcytoskeletal proteins are found in the plasma membranesurrounding the extruded nucleus [22,29–31]. Temporalsynthesis and degradation of individual membrane andcytoskeletal proteins are thought to play crucial roles inthe reorganization process [30].

Questions remain regarding the dynamic aspects ofmembrane and cytoskeletal protein reorganizations dur-ing erythropoiesis and little is known of the contribution of

411

Review Trends in Cell Biology July 2011, Vol. 21, No. 7

these changes to enucleation. These reorganizations mightfacilitate the enucleation process because morphologicalstudies imply that the nucleus in the late stages of eryth-ropoiesis moves to a patch of membrane that is deficient inthe spectrin network [29]. However, mice bearing muta-tions in or deletions of a-spectrin, b-spectrin, ankyrin,band 3, protein 4.1 or protein 4.2 do not exhibit significantincreases in circulating nucleated red blood cells [32–36].In addition, fetal liver erythroblasts purified from thesemice show no obvious defects of enucleation in in vitroculture (P. Ji et al. unpublished data). It remains to bedetermined whether or not a deficiency of any of thesemembrane or cytoskeletal proteins, individually or togeth-er, causes a drop in the rate or extent of enucleation.

Earlier studies using mammalian erythroid cell linesshowed that the intermediate filament vimentin is absentor becomes rapidly lost during maturation [37,38]. Bycontrast, in other vertebrates intermediate filaments ap-pear to be important in anchoring the condensed nucleusand preventing it from being extruded [39,40]. Hypotheti-cally, depletion of vimentin or other intermediate filamentproteins could contribute to nuclear displacement but notbe sufficient for mammalian erythroblast enucleation. It isnot clear whether other cytoskeleton proteins might beinvolved in this complex process.

There is still debate as to precisely how the reticulocyteand nucleus undergo final separation. The findings notedabove concerning the cortical actin ring suggest a processakin to cytokinesis during asymmetric cell division. How-ever, early electron micrographic studies revealed multiplevesicles forming in the region between the extruding nu-cleus and incipient reticulocyte [22]. Recent evidence sug-gests that endocytic vesicle trafficking is essential forenucleation. Inhibition of clathrin-dependent vesicle traf-ficking blocked enucleation of primary fetal liver erythro-blasts without affecting normal differentiation orproliferation [41]. Clearly, much needs to be learned aboutthe role of cytoskeletal proteins and membranes in thisprocess.

Erythropoietic microenvironment and enucleationStudies in the 1970s showed that the extruded nucleus isenveloped by a portion of the plasma membrane withdistinctive cell surface proteins different from those onthe reticulocyte [29,42]. The exoplasmic leaflet of the mem-brane enveloping the nucleus contains high levels of phos-phatidylserine, which serves as a signal for macrophageengulfment; masking phosphatidylserine on the nucleusprevents macrophage phagocytosis of the extruded nuclei[17]. Correspondingly, the cell surface of mature red bloodcells contains CD47, which serves as a ‘self’ marker andprevents elimination by macrophages [43].

Macrophages also play important roles in the develop-ment of erythroid cells. Erythropoiesis in fetal liver andspleen and adult bone marrow is regulated by its microen-vironment, a domain known as the ‘erythroblastic island’(see review [44]). Morphologically, the island containscentral macrophages surrounded by a ring of erythro-blasts. The central macrophages are also called nursingcells, in that they provide iron and developmental signalsto the surrounding erythroblasts [45].

412

Many cell surface proteins have been implicated in themacrophage–erythroblast interaction. Among these, eryth-roblast macrophage protein (Emp) seems to be particularlyimportant for erythroblast enucleation. Emp is normallyexpressed on the cell surface of both macrophages anderythroblasts [46,47]. Genetic ablation of Emp in themouse leads to a dramatic increase of nucleated red bloodcells in the peripheral blood, and no fetal liver erythroblas-tic islands are observed in Emp-deficient mice [48]. Impor-tantly, a deficiency of Emp leads to a disruption of actinlocalization during terminal erythroblast differentiation[48]. It would be interesting to determine whether Empplays any role in the Rac GTPases–mDia2 or other path-ways in enucleation. However, primary erythroid progeni-tors do undergo differentiation and enucleation whencultured in the absence of macrophages or other non-erythroid cells [20].

Other proteins that are involved in the macrophage–

erythroblast interaction include the tumor suppressor ret-inoblastoma (Rb) protein. Rb deficiency in the mouse fetalliver somehow prevents interactions between macro-phages and erythroblasts and blocks erythroblast enucle-ation [49]. This is probably owing to the fact that loss of Rbblocks normal macrophage differentiation because wildtype macrophages can bind to Rb-deficient erythroblastsand induce their enucleation [49]. However, Rb also plays acell-intrinsic role in erythropoiesis because loss of Rbblocks erythroblast differentiation at an early stage [50].Clearly, we need to learn more about the function of Rb andthe genes it regulates in both erythroblasts andmacrophages.

Mammalian erythroblasts also interact with other cel-lular and matrix components during development. In vitrostudies indicated that enucleated murine erythroleukemiacells are significantly increased when cultured in fibronec-tin-coated dishes [51]. Fibronectin, through interactionswith a4b1 integrin on the erythroblast cell surface, is alsoimportant for the survival of the late-stage erythroblast bypreventing apoptosis [1]. How these interactions affectenucleation is unclear.

During embryogenesis, prior to ‘definitive’ erythropoie-sis in the fetal liver, red cell formation begins in the yolksac with production of ‘primitive’ erythrocytes that expressembryonic globins and that retain their nuclei in theembryonic circulation [52]. It was long believed that circu-lating primitive erythroid cells never lose their nuclei andthus share morphological similarities with those of othervertebrates [53]. However, a recent report demonstratedthat primitive mammalian erythroid cells do undergo enu-cleation; in contrast to definitive erythroid cells in the fetalliver, terminal differentiation and enucleation of primitiveerythroid cells take place in the circulation [54]. This raisesquestions concerning how nucleated circulating primitiveerythroid cells maintain their interactions with themicroenvironment.

MicroRNAs in enucleationMicroRNAs are a class of small RNAs that downregulateexpression of specific target genes post-transcriptionally[55]. They play a wide variety of roles in hematopoiesis[56,57], specifically in erythropoiesis [58]. Among these

Review Trends in Cell Biology July 2011, Vol. 21, No. 7

microRNAs, the miR-144/451 cluster is directly activatedby the erythroid-important GATA1 transcription factorand is required for erythropoiesis [59]. Specifically, inZebrafish miR-144 directly suppresses the levels of Klfd,an erythroid specific Kruppel-like transcription factor, toregulate a-hemoglobin synthesis [60]. By contrast, miR-451 targets GATA2 mRNA to promote Zebrafish erythroidcell maturation [61]. Recent data from the miR-144/451knockout mouse revealed that murine erythrocytes withmiR-144/451 deficiency exhibit a cell autonomous im-pairment of late erythroblast maturation and increasedsusceptibility to oxidant damage [62].

By contrast, over-expression of miR-191, a microRNAthat is normally down-regulated during erythropoiesis,blocks chromatin condensation and enucleation with mi-nor effects on proliferation and differentiation [63]. Riok3and Mxi1, two erythroid enriched and developmentally up-regulated genes, were found to be principal downstreamtargets of miR-191. Knocking down either of these twoproteins mimicked the effects of miR-191 over-expression,blocking both chromatin condensation and enucleation. Inthis aspect, it is interesting to note that Mxi1 is an antago-nist of c-Myc [64], whose level we noted earlier graduallydecreases during erythropoiesis [19]. In addition, knock-down of either Mxi1 or Riok3 blocked Gcn5 down-regula-tion in erythroblasts [63]. This and other studies illustratethe importance of the miR-191–Mxi1–c-myc–Gcn5 andmiR-191–Riok3–Gcn5 pathways in the regulation of termi-nal erythroid cell chromatin condensation (Figure 1). In-deed, the significance of this pathway needs to be furtherinvestigated, as do the roles of many other abundantlyexpressed microRNAs in erythropoiesis.

Concluding remarksEnucleation of mammalian erythroblasts is a complexprocess that involves a series of morphological and struc-tural changes. In this review, we summarized currentprogress in understanding mammalian erythroid cell enu-cleation. Most of these studies utilized in vitro erythroblastculture systems because the extruded nucleus is releasedinto the culture medium and not immediately engulfed bymacrophages. The Emp-deficient mouse might provide anintriguing system to study enucleation in vivo. However, itis not clear whether Emp plays a specific role in enucle-ation because the Emp-deficient mouse has significantamounts of immature proerythroblasts [48], which indi-cates that Emp is also critical for early stages of erythro-poiesis. Similar to Emp, Rac GTPases are also importantfor the early stage of erythropoiesis [65], in addition totheir functions in enucleation [28]. These facts raise diffi-culties in associating proteins with chromatin condensa-tion or enucleation per se because any inhibition of earlystages of erythroblast differentiation will probably reduceformation of enucleated reticulocytes.

Furthermore, possible lethality associated with enucle-ation failure and the redundancy of proteins required forkey steps in this complex pathway add to the dilemma ofin vivo investigations. In this respect, it would be desirableto establish inducible knockout mouse models to studyenucleation. Examples would be mice in which a gene isinactivated solely in late erythroblasts. Significant

numbers of circulating nucleated red blood cells are seenwhen the human body is under chronic stress or in certainpathological processes [66]. Studies of available mousemodels in these conditions might provide informativeinsights into the enucleation process.

Enhancing erythroblast enucleation has a practical sig-nificance clinically in producing red blood cells in culturefor transfusion. Current practice in transfusion medicinefaces constant difficulties in obtaining sufficient supplies ofspecific red blood cell subtypes [67]. Recently, ex vivo orin vitro production of human red blood cells from hemato-poietic stem cells, embryonic stem cell and induced plurip-otent stem cells have been extensively explored [68–72]. Inaddition to the challenges of generating sufficient amountof cells for transfusion purposes, the reduced ability ofthese cultured cells to fully enucleate remains an obstacle.Future endeavors focusing on the efficiency of enucleationin late erythroblasts derived in culture from various typesof stem or progenitor cells, as well as the understanding ofthe basic biology of red cell enucleation, will help in trans-ferring these studies from bench to bedside.

Clearly other major questions concerning enucleationneed to be addressed. First, the significance of the ery-throid microenvironment, including macrophages, in enu-cleation needs to be further investigated. For example, aswe discussed above, Emp deficiency might affect earlystages of erythropoiesis, which in turn are required forlater enucleation. Second, we do not know whether or howchromatin condensation triggers or facilitates the enucle-ation process. As an example, could chromatin condensa-tion or a modified nuclear envelope induce the RacGTPases–mDia2 pathway to promote enucleation(Figure 1)? Third, why is enucleation only seen in mam-mals? In ‘lower’ vertebrates, intermediate filaments pre-vent enucleation by anchoring the nucleus. Are othersubcellular structures also involved? Could loss of expres-sion of these genes in erythroid cells during vertebrateevolution promote enucleation? Comparative genomic orproteomic studies across species would provide clues.

AcknowledgmentsThis study was supported by NIH grant P01 HL 32262, a research grantfrom Amgen (to H.F.L.) and by intramural funds from the Temasek LifeSciences Laboratory to M.M.-H. P.J. is the recipient of a Leukemia andLymphoma Society fellowship and a National Institutes of Health (NIH)Pathway to Independence Award (K99/R00).

References1 Eshghi, S. et al. (2007) Alpha4beta1 integrin and erythropoietin

mediate temporally distinct steps in erythropoiesis: integrins in redcell development. J. Cell Biol. 177, 871–880

2 Zhang, J. et al. (2003) Role of Ras signaling in erythroid differentiationof mouse fetal liver cells: functional analysis by a flow cytometry-basednovel culture system. Blood 102, 3938–3946

3 Neelin, J.M. et al. (1964) The histones of chicken erythrocyte nuclei.Can. J. Biochem. Physiol. 42, 1743–1752

4 Hnilica, L.S. (1964) The specificity of histones in chicken erythrocytes.Experientia 20, 13–14

5 Miki, B.L. and Neelin, J.M. (1975) The histones of rainbow trouterythrocytes include an erythrocyte-specific histone. Can. J.Biochem. 53, 1158–1169

6 Awai, M. et al. (1968) Studies on the mechanism of denucleation of theerythroblast. Acta Haematol. 39, 193–202

7 Muir, A.R. and Kerr, D.N. (1958) Erythropoiesis: an electronmicroscopical study. Q. J. Exp. Physiol. Cogn. Med. Sci. 43, 106–114

413

Review Trends in Cell Biology July 2011, Vol. 21, No. 7

8 Orlic, D. et al. (1965) An ultrastructural study of erythropoietin-induced red cell formation in mouse spleen. J. Ultrastruct. Res. 13,516–542

9 Simpson, C.F. and Kling, J.M. (1967) The mechanism of denucleationin circulating erythroblasts. J. Cell Biol. 35, 237–245

10 Skutelsky, E. and Danon, D. (1967) An electron microscopic study ofnuclear elimination from the late erythroblast. J. Cell Biol. 33, 625–635

11 Skutelsky, E. and Danon, D. (1970) Comparative study of nuclearexpulsion from the late erythroblast and cytokinesis. Exp. Cell Res.60, 427–436

12 Testa, U. (2004) Apoptotic mechanisms in the control of erythropoiesis.Leukemia 18, 1176–1199

13 Ishizaki, Y. et al. (1998) A role for caspases in lens fiber differentiation.J. Cell Biol. 140, 153–158

14 Weil, M. et al. (1999) Caspase activation in the terminal differentiationof human epidermal keratinocytes. Curr. Biol. 9, 361–364

15 Carlile, G.W. et al. (2004) Caspase-3 has a nonapoptotic function inerythroid maturation. Blood 103, 4310–4316

16 Kolbus, A. et al. (2002) Raf-1 antagonizes erythroid differentiation byrestraining caspase activation. J. Exp. Med. 196, 1347–1353

17 Yoshida, H. et al. (2005) Phosphatidylserine-dependent engulfment bymacrophages of nuclei from erythroid precursor cells. Nature 437,754–758

18 Krauss, S.W. et al. (2005) Nuclear substructure reorganization duringlate-stage erythropoiesis is selective and does not involve caspasecleavage of major nuclear substructural proteins. Blood 106, 2200–2205

19 Jayapal, S.R. et al. (2010) Down-regulation of Myc is essential forterminal erythroid maturation. J. Biol. Chem. 285, 40252–40265

20 Ji, P. et al. (2010) Histone deacetylase 2 is required for chromatincondensation and subsequent enucleation of cultured mouse fetalerythroblasts. Haematologica 95, 2013–2021

21 Popova, E.Y. et al. (2009) Chromatin condensation in terminallydifferentiating mouse erythroblasts does not involve specialarchitectural proteins but depends on histone deacetylation.Chromosome Res. 17, 47–64

22 Koury, S.T. et al. (1989) Cytoskeletal distribution and function duringthe maturation and enucleation of mammalian erythroblasts. J. CellBiol. 109, 3005–3013

23 Chasis, J.A. et al. (1989) Membrane assembly and remodeling duringreticulocyte maturation. Blood 74, 1112–1120

24 Byers, T.J. and Branton, D. (1985) Visualization of the proteinassociations in the erythrocyte membrane skeleton. Proc. Natl.Acad. Sci. U.S.A. 82, 6153–6157

25 Gratzer, W.B. (1981) The red cell membrane and its cytoskeleton.Biochem. J. 198, 1–8

26 Tse, W.T. and Lux, S.E. (1999) Red blood cell membrane disorders.Br. J. Haematol. 104, 2–13

27 Takano-Ohmuro, H. et al. (1996) Distribution of actin, myosin, andspectrin during enucleation in erythroid cells of hamster embryo. CellMotil. Cytoskeleton 34, 95–107

28 Ji, P. et al. (2008) Enucleation of cultured mouse fetal erythroblastsrequires Rac GTPases and mDia2. Nat. Cell Biol. 10, 314–321

29 Geiduschek, J.B. and Singer, S.J. (1979) Molecular changes in themembranes of mouse erythroid cells accompanying differentiation. Cell16, 149–163

30 Wickrema, A. et al. (1994) Changes in cytoskeletal proteins and theirmRNAs during maturation of human erythroid progenitor cells. J. CellPhysiol. 160, 417–426

31 Lee, J.C. et al. (2004) Mechanism of protein sorting duringerythroblast enucleation: role of cytoskeletal connectivity. Blood103, 1912–1919

32 Wandersee, N.J. et al. (2001) Defective spectrin integrity and neonatalthrombosis in the first mouse model for severe hereditaryelliptocytosis. Blood 97, 543–550

33 Yi, S.J. et al. (1997) Red cell membranes of ankyrin-deficient nb/nb micelack band 3 tetramers but contain normal membrane skeletons.Biochemistry 36, 9596–9604

34 Peters, L.L. et al. (1996) Anion exchanger 1 (band 3) is required toprevent erythrocyte membrane surface loss but not to form themembrane skeleton. Cell 86, 917–927

35 Shi, Z.T. et al. (1999) Protein 4.1R-deficient mice are viable buthave erythroid membrane skeleton abnormalities. J. Clin. Invest.103, 331–340

414

36 Peters, L.L. et al. (1999) Mild spherocytosis and altered red cell iontransport in protein 4.2-null mice. J. Clin. Invest. 103, 1527–1537

37 Dellagi, K. et al. (1983) Alteration of vimentin intermediate filamentexpression during differentiation of human hemopoietic cells. EMBO J.2, 1509–1514

38 Ngai, J. et al. (1984) Differentiation of murine erythroleukemia cellsresults in the rapid repression of vimentin gene expression. J. Cell Biol.99, 306–314

39 Granger, B.L. et al. (1982) Synemin and vimentin are components ofintermediate filaments in avian erythrocytes. J. Cell Biol. 92, 299–312

40 Virtanen, I. et al. (1979) Nucleus-anchoring cytoskeleton in chicken redblood cells. Cell Biol. Int. Rep. 3, 157–162

41 Keerthivasan, G. et al. (2010) Vesicle trafficking plays a novel role inerythroblast enucleation. Blood 116, 3331–3340

42 Skutelsky, E. and Farquhar, M.G. (1976) Variations in distribution ofcon A receptor sites and anionic groups during red blood celldifferentiation in the rat. J. Cell Biol. 71, 218–231

43 Oldenborg, P.A. et al. (2000) Role of CD47 as a marker of self on redblood cells. Science 288, 2051–2054

44 Chasis, J.A. and Mohandas, N. (2008) Erythroblastic islands: niches forerythropoiesis. Blood 112, 470–478

45 Manwani, D. and Bieker, J.J. (2008) The erythroblastic island. Curr.Top. Dev. Biol. 82, 23–53

46 Hanspal, M. and Hanspal, J.S. (1994) The association of erythroblastswith macrophages promotes erythroid proliferation and maturation: a30-kD heparin-binding protein is involved in this contact. Blood 84,3494–3504

47 Hanspal, M. et al. (1998) Molecular identification and functionalcharacterization of a novel protein that mediates the attachment oferythroblasts to macrophages. Blood 92, 2940–2950

48 Soni, S. et al. (2006) Absence of erythroblast macrophage protein (Emp)leads to failure of erythroblast nuclear extrusion. J. Biol. Chem. 281,20181–20189

49 Iavarone, A. et al. (2004) Retinoblastoma promotes definitiveerythropoiesis by repressing Id2 in fetal liver macrophages. Nature432, 1040–1045

50 Sankaran, V.G. et al. (2008) Rb intrinsically promotes erythropoiesis bycoupling cell cycle exit with mitochondrial biogenesis. Genes Dev. 22,463–475

51 Patel, V.P. and Lodish, H.F. (1987) A fibronectin matrix is requiredfor differentiation of murine erythroleukemia cells into reticulocytes.J. Cell Biol. 105, 3105–3118

52 Palis, J. and Yoder, M.C. (2001) Yolk-sac hematopoiesis: the first bloodcells of mouse and man. Exp. Hematol. 29, 927–936

53 Lazarides, E. (1987) From genes to structural morphogenesis: thegenesis and epigenesis of a red blood cell. Cell 51, 345–356

54 Kingsley, P.D. et al. (2004) Yolk sac-derived primitive erythroblastsenucleate during mammalian embryogenesis. Blood 104, 19–25

55 Bartel, D.P. (2009) MicroRNAs: target recognition and regulatoryfunctions. Cell 136, 215–233

56 Chen, C.Z. et al. (2004) MicroRNAs modulate hematopoietic lineagedifferentiation. Science 303, 83–86

57 Chen, C.Z. and Lodish, H.F. (2005) MicroRNAs as regulators ofmammalian hematopoiesis. Semin. Immunol. 17, 155–165

58 Zhao, G. et al. (2010) MicroRNAs in erythropoiesis. Curr. Opin.Hematol. 17, 155–162

59 Dore, L.C. et al. (2008) A GATA-1-regulated microRNA locus essentialfor erythropoiesis. Proc. Natl. Acad. Sci. U.S.A. 105, 3333–3338

60 Fu, Y.F. et al. (2009) Mir-144 selectively regulates embryonic alpha-hemoglobin synthesis during primitive erythropoiesis. Blood 113,1340–1349

61 Pase, L. et al. (2009) miR-451 regulates zebrafish erythroid maturationin vivo via its target gata2. Blood 113, 1794–1804

62 Rasmussen, K.D. et al. (2010) The miR-144/451 locus is required forerythroid homeostasis. J. Exp. Med. 207, 1351–1358

63 Zhang, L. et al. (2011) miR-191 regulates mouse erythroblast enucleationby down-regulating Riok3 and Mxi1. Genes Dev. 25, 119–124

64 Schreiber-Agus, N. et al. (1995) An amino-terminal domain of Mxi1mediates anti-Myc oncogenic activity and interacts with a homolog ofthe yeast transcriptional repressor SIN3. Cell 80, 777–786

65 Kalfa, T.A. et al. (2010) Rac1 and Rac2 GTPases are necessary for earlyerythropoietic expansion in the bone marrow but not in the spleen.Haematologica 95, 27–35

Review Trends in Cell Biology July 2011, Vol. 21, No. 7

66 Hermansen, M.C. (2001) Nucleated red blood cells in the fetus andnewborn. Arch. Dis. Child. Fetal Neonatal Ed. 84, F211–F215

67 Douay, L. and Andreu, G. (2007) Ex vivo production of human red bloodcells from hematopoietic stem cells: what is the future in transfusion?Transfus. Med. Rev. 21, 91–100

68 Panzenbock, B. et al. (1998) Growth and differentiation of human stemcell factor/erythropoietin-dependent erythroid progenitor cells in vitro.Blood 92, 3658–3668

69 Giarratana, M.C. et al. (2005) Ex vivo generation of fully maturehuman red blood cells from hematopoietic stem cells. Nat.Biotechnol. 23, 69–74

70 Anstee, D.J. (2010) Production of erythroid cells from humanembryonic stem cells (hESC) and human induced pluripotent stemcells (hiPS). Transfus. Clin. Biol. 17, 104–109

71 Fujimi, A. et al. (2008) Ex vivo large-scale generation of human redblood cells from cord blood CD34+ cells by co-culturing withmacrophages. Int. J. Hematol. 87, 339–350

72 Lu, S.J. et al. (2008) Biologic properties and enucleation of red bloodcells from human embryonic stem cells. Blood 112, 4475–4484

73 Snyder, G.K. and Sheafor, B.A. (1999) Red blood cells: centerpiecein the evolution of the vertebrate circulatory system. Am. Zool. 39,189–198

415