5-Azacytidine reactivates the erythroid differentiation potentialof the myeloid-restricted murine cell line 32D Ro

Marta Baiocchi,a Cristina Di Rico,b Roberta Di Pietro,c Angela Di Baldassarre,cand Anna Rita Migliacciob,c,*

a Department of Hematology, Instituto Superiore Sanita, Rome, Italyb Department of Clinical Biochemistry, Istituto Superiore Sanita, Rome, Italy

c Biomorphology, University G. D’Annunzio, Chieti, Italy

Received 4 September 2002, revised version received 25 November 2002

Abstract

32D cells grown for 1 year in interleukin-3 (IL-3) and granulocyte colony-stimulating factor (G-CSF) generated the 32D Ro cell linewhich retained the parental mast cell phenotype but lost ability to generate erythroid cells in response to erythropoietin (EPO). In order toclarify the mechanisms underlying such restriction, we compared 32D and 32D Ro cells for their capacity to express erythroid-specifictranscription factors (Gata1, Gata2, Scl, Nef2, Eklf, and Id) and the capacity of short exposure to 5-azacytidine (5-AzaC) to reactivateerythroid differentiation potential in 32D Ro cells. By Northern analysis, the two cell lines expressed similar levels of all these genes.However, after being treated with 5-AzaC, 32D Ro cells acquired the ability to generate EPO-dependent clones (1 clone/104 cells) whichgave rise to EPO-dependent cell lines. All the 10 EPO-responsive cell lines independently isolated from 5-AzaC-treated 32D Ro cells haderythroid morphology and expressed high levels of !- and "-globin. In contrast, none of the IL-3-dependent and granulocyte/macrophagecolony-stimulating factor-dependent clones concurrently isolated, as a control, showed erythroid properties. Therefore, 5-AzaC treatmentreactivates the potential of the myeloid-restricted 32D Ro cells to generate EPO-responsive erythroid clones suggesting that genemethylation played an important role in the G-CSF-mediated restriction/activation of the differentiation potential of these cells.© 2003 Elsevier Science (USA). All rights reserved.

Keywords: 5-Azacytidine; Methylation; Erythroid differentiation; 32D cell line; Stem cell plasticity

Introduction

Hematopoiesis begins at the level of a multipotent he-matopoietic stem cell that undergoes commitment generat-ing a series of cells progressively restricted in their differ-entiation potential [1,2]. One of the steps in this orderly andunidirectional sequence of events is the generation, from acommon multipotent progenitor cell, of two cells, one re-stricted for the myelo-monocytic and the other for the eryth-roid-megakaryocytic lineage [3]. This process has been as-sociated with a series of epigenetic modifications of the

cellular genome such as activation of the expression oflineage-specific transcription factors [4], reorganization ofthe chromosomes into lineage-specific nucleosomal struc-tures [5], and de novo DNA methylation [6] and is associ-ated with the cell’s ability to respond to a sequence ofspecific growth factors [7]. The common multipotent pro-genitor cell is dependent for growth on the presence of stemcell factor (SCF)1 and interleukin-3 (IL-3). Later on, re-stricted myelo-monocytic and erythroid-megakaryocyticprogenitor cells show the additional requirement for eithergranulocyte/macrophage colony-stimulating factor (GM-CSF), macrophage (M)-CSF, and granulocyte (G)-CSF or

* Corresponding author. Laboratorio Biochimica Clinica, Istituto Su-periore Sanita, Viale Regina Elena 299, 00161, Rome, Italy. Fax: !0039-06-49387143.

E-mail address: migliar@iss.it (A.R. Migliaccio).

1 Proteins and genes have been abbreviated according to the GeneticNomenclature Guide (Trends in Genetics, A. Stewart, (Ed.), Elsevier Sci-ence Publisher, Cambridge, UK, 1995).

R

Available online at www.sciencedirect.com

Experimental Cell Research 285 (2003) 258–267 www.elsevier.com/locate/yexcr

0014-4827/03/$ – see front matter © 2003 Elsevier Science (USA). All rights reserved.doi:10.1016/S0014-4827(03)00032-6

erythropoietin (EPO) and thrombopoietin (TPO), respectively[7]. The exact relationship between epigenetic DNA modifi-cations and growth factor response has not yet been clarified.Normal stem/progenitor cells have been used to identify

genes expressed during lineage commitment [3,8] whileembryonic stem cell lines or IL-3-dependent cell lines (suchas 32D and FDCP-mix) [9,10] have been used to identifygenes functionally involved in the commitment process.The IL-3-dependent cell lines, in particular, generate at lowfrequency (1 clone/103–104 cells) cells that respond to EPO,GM-CSF, or G-CSF [11] and that differentiate in the pres-ence of these growth factors along the appropriate hemo-poietic lineage. Therefore, although the majority of thesecells have a mast cell morphology, some of them representa cellular model for the common multipotent progenitorcell. 32D cells cultured for more than a year in the presenceof both IL-3 and G-CSF originated the myeloid-restrictedcell line 32D Ro [12] which had the same phenotype of theparental cell line but had lost its potential to respond to EPOand TPO [11 and unpublished observation] while respond-ing more readily to GM-CSF and G-CSF [13 and thismanuscript]. For this reason, 32D Ro cells are considered acellular model for the myeloid-restricted progenitor cell[14]. It is not clear, as yet, what the molecular mechanismwas that occurred during the long exposure of 32D cells toG-CSF that restricted their diffrentiation potential.5-Azacytidine (5-AzaC) is a nucleoside analog which

competes with cytidine for insertion into the DNA during itsreplication [6]. Its chemical structure is such that it does notinhibit normal DNA function with the exception of de novomethylation after replication. Therefore, it interferes with allof those gene functions that are methylation dependent andhas been often used to prove involvement of DNA methyl-ation in cell differentiation, as pioneered in the study byTaylor and Jones [15] that led to the discovery of the firstmaster regulatory gene, myoD [16].To clarify the nature of the molecular event induced by

G-CSF and that restricted the differentiation potential of32D cells, we compared the levels of expression of eryth-roid-specific transcription factors in the two cell lines anddetermined the ability of 5-AzaC to reactivate the erythroidpotential of 32D Ro cells. While the two cell lines did notshow significant differences in gene expression, 5-AzaCtreatment did induce the 32D Ro cell line to respond to EPOand to generate erythroid clones. These results suggest thatDNA methylation was involved in the G-CSF-induced lin-eage restriction of 32D Ro cells.

Materials and methods

Hematopoietic growth factors

Pure recombinant murine IL-3 and GM-CSF were gentlyprovided by Dr. Clive Wood (Genetics Institute Cambridge,MA, USA). Pure recombinant human EPO and G-CSF were

provided by Drs. J. Egrie and L. Souza, respectively (AmgenThousand Oaks, CA, USA). The growth factors were used atconcentrations that induced maximal number of colonies inserum-deprived cultures of normal murine bone marrow (i.e.,100 units/mL for IL-3, 10 units/mL for GM-CSF, 1000units/mL for G-CSF, 1 unit/mL for EPO) [17,18].

Cell lines

The 32D and 32D Ro cell lines were maintained inmodified McCoys medium (GIBCO-BRL, Grand Island,NY) supplemented with 10% fetal bovine serum (FBS),horse serum (HS), L-glutamine (2 mM), antibiotics, andIL-3 as described [11]. The erythroid subclone of 32D (32DEpo) was maintained under the same conditions but in thepresence of EPO (11). Cultures were incubated at 35°C in a5% CO2 atmosphere. The experimental design for the incu-bation of 32D Ro cells with 5-AzaC is described in Fig. 1.Briefly, 32D Ro cells were seeded in Iscove’s modifiedDulbecco’s medium (IMDM; GIBCO-BRL) (105 cells/mL)under serum-deprived conditions and stimulated with IL-3(100 units/mL) in the absence or in the presence of increas-ing concentrations of 5-AzaC (1, 3, or 10 #M). After 24–48h of incubation, cells were washed twice with IMDM andcounted with a hemocytometer. The frequency of cells ca-pable of growing either in the absence of growth factors(negative control) or in the presence of IL-3 (positive con-trol), EPO, or GM-CSF was evaluated in semisolid medium(0.8% final concentration of methylcellulose) under serum-deprived conditions. After 10 days of incubation at 37°C in5% CO2, the number of colonies present in the dishes wasdetermined by eye with the help of an inverted microscope.Single colonies were harvested with a thin glass capillarytube and divided into two aliquots. One aliquot was used forgene expression analysis by reverse-transcriptase polymer-ase chain reaction while the other one was expanded inmicroplates (one colony/well in a 24-multiwell system) con-taining McCoys medium (1 mL) and the same growth factorwhich was used for the selection. Clonal cell lines were thengenerated by transferring the cells grown into each well intoindividual flasks which were then regularly passed twice aweek. A clone was considered as a line after at least 10passages (5 weeks of culture). Gene expression analysis wasperformed first when the single clones were harvested fromthe semisolid dish, then when they were expanded from thewell to the flask, and finally after 5 weeks of regular pas-sages in liquid culture. Gene expression was analyzed byRT-PCR in the first two cases and by Northern analysis inthe last case. All the chemicals were obtained from Sigma(St. Louis, MO, USA), unless otherwise indicated.

RNA extraction

RNA was extracted from 104–106 cells as described [19].Yeast tRNA (Sigma) was added as a carrier before theextraction from a low number of cells.

259M. Baiocchi et al. / Experimental Cell Research 285 (2003) 258–267

Reverse-transcriptase polymerase chain reaction(RT-PCR)

RT-PCR was performed with the Gene Amp RNA PCRKit (Perkin–Elmer, Norwalk, CT) according to the manifac-turer’s instructions. Briefly, cDNA was retrotranscribed at42°C for 30 h using oligo dT as a primer. Aliquots of cDNA

retrotranscribed from RNA corresponding to 103 cells wereused in each PCR reaction and all of the genes analyzedwere amplified from the same cDNA preparation. The am-plification reactions were performed with the Cetus DNAThermal Cycler (Perkin–Elmer) using the couple of primersand the PCR conditions already described [20,21]. To min-imize the effects of DNA contaminations, each pair of

Fig. 1. Experimental design of the treatment with 5-AzaC of the 32D Ro cell line. See Materials and Methods for details.

260 M. Baiocchi et al. / Experimental Cell Research 285 (2003) 258–267

primers was chosen in regions of the genes separated by atleast one intron. rRNA and mock RNA extracted from wellscontaining culture medium but no cells were used as nega-tive controls. The amplified fragments were separated byelectrophoresis on agarose gel which was then blotted ontoZeta probe membranes (BioRad, Richmond, CA, USA) andprobed with 35-mer primers internal to the amplified se-quences end-labeled with the 5" DNA Terminus LabelingSystem (GIBCO-BRL). Membranes were exposed for 4 hwith Kodak X-Omat AR films (Sigma). All the procedureswere according to standard protocols [22].

Northern blot analysis

RNA was size fractionated by electrophoresis on agarose(1%, w/v) gel under denaturing conditions and blotted ontonylon membranes (BioRad) that were subsequently hybrid-ized with Scl [23] (a gift from Dr. G. Begley), Id1 [24] (a

gift from Dr. H. Weintraub), Gata1 [25] and Gata2 [26](both a gift from Dr. S. Orkin), Eklf [27] (a gift from Dr. J.Bieker), Nfe2 [28] (a gift from Dr. Y.W. Kan), the eryth-ropoietin receptor (EpoR) [29] (a gift from Dr. A.D’Andrea), !-globin, or "-globin [30]. Each probe wasradiolabeled by random oligonucleotide priming(Boheringer Mannheim, Germany) to a specific activity of4–8 # 108 dpm/mg. After probing, the membranes werewashed as recommended by the manufactures and exposedfor the appropriate lengths of time with X Omat film (Sig-ma) in cassettes for autoradiography (Amersham). All theprocedures were carried out according to standard protocols[22].

Cell morphology

Cell morphology was analyzed according to standardcriteria on cytocentrifuged (Shandon, Astmoor, UK) slides

Fig. 2. (A) Cloning efficiency of 32D (white bars) and 32D Ro (black bars) cells in serum-deprived cultures in the absence of growth factors (None) or inthe presence of IL-3, EPO, GM-CSF, or G-CSF, as indicated. The results are presented as the means ($SD) of at least three separate experiments performedin triplicate. (B) Northern analysis for the expression of erythroid genes (Gata1, Gata2, Scl, Nfe2, Eklf, Id, and "- and !-globin) in 32D and 32D Ro cells.The ethidium bromide staining of the gel before blotting is shown as control of the rRNA loaded in each lane.

261M. Baiocchi et al. / Experimental Cell Research 285 (2003) 258–267

stained with May–Grunwald Giemsa. Hemoglobin-contain-ing cells were identified by benzidine staining [31] and cellviability was assessed by trypan blue exclusion.

Results

Comparison of the erythroid differentiation potential of32D and 32D Ro cells

Under serum-deprived conditions, 32D and 32D Ro cellsdid not grow in the absence of growth factors and clonedwith a comparable efficiency of %10% when stimulated byIL-3 (Fig. 2A). The two cell lines, however, differed widelyin their response to erythroid (EPO) or myeloid (G-CSF andGM-CSF) growth factors. 32D cells cloned with a similarlylow efficiency (1 clone every 103 cells) in cultures stimu-lated with EPO, GM-CSF, or G-CSF while 32D Ro cellsnever grew in the presence of EPO and cloned readily (0.5%cloning efficiency) in GM-CSF- or G-CSF-stimulated cul-tures.The levels of expression of transcription factors thought

to be important for erythroid differentiation in the two linesare presented in Fig. 2B. No difference was observed be-tween 32D and 32D Ro cells in the levels of expression ofGata1 (barely expressed by both), Gata2, Scl, Nfe-2, and Id.In contrast, !-globin was expressed at levels detectable byNorthern only in 32D cells while expression of Eklf wasdetected mainly in 32D Ro cells.

Incubation with 5-AzaC activates the ability of 32D Rocells to respond to EPO

The effects of increasing concentrations of 5-AzaC onthe proliferation of 32D Ro cells in liquid cultures stimu-lated with IL-3 and in semisolid cultures (cloning effi-ciency) stimulated with either IL-3 or EPO are presented inFig. 3. When transferred into fresh medium, 32D Ro cellsstarted to proliferate immediately with an apparent doublingtime of 24 h over 2 days (Fig. 3, top panel). In contrast, if5-AzaC had been added to the medium, the number of cellsremained constant for the first 24 h and started to increaseonly after 48 h.Short exposure (1–2 days) to fresh medium consistently

increased the number of 32D Ro cells which formed colo-nies in response to IL-3 (from 10 to 40% of cloning effi-ciency) (Fig. 2, middle panel). On the other hand, if 5-AzaCwas present, 32D Ro cells cloned in IL-3 with a constantefficiency of 10% with the exception of cells that had beenexposed for 2 days at the lowest 5-AzaC concentration thatcloned with an efficiency of 40%. These results indicate thatexposure of 32D Ro cells to 5-AzaC had no major toxiceffects with the exception of a delay in their proliferation forthe first 24 h.

Untreated 32D Ro cells did not clone in response to EPOeven when exposed to fresh medium (Fig. 3, lower panel).In contrast, 0.1–1% of the cells that had been exposed to5-AzaC cloned in response to EPO. The number of EPO-sensitive cells was concentration dependent since the high-est number was observed when the cells had been exposedto 3 #M 5-AzaC. Therefore, short-term exposure to 5-AzaCreactivates the capacity of 32D Ro cells to clone in responseto EPO.

Fig. 3. Time course analysis of the generation of EPO-responsive cellsfrom 32D Ro cells treated with increasing concentrations of 5-AzaC (0–10#M, as indicated). The total number of cells observed during the treatmentand the number of IL-3-responsive (as control of toxicity) and of EPO-responsive clones present among the treated cells are presented in theupper, middle, and lower panels, respectively.

262 M. Baiocchi et al. / Experimental Cell Research 285 (2003) 258–267

EPO-responsive clones obtained from 5-AzaC-treated32D Ro cells express high levels of "-globin

To identify the erythroid-differentiation potential ofEPO-responsive clones obtained from 5-AzaC-treated 32DRo cells, colonies were randomly harvested from the dishand divided into two aliquots. The first aliquot was used toanalyze the level of "-globin expression by RT-PCR (Fig.4) while the second one was passed into liquid culture todetermine for how long the response to EPO would betransmitted to the progeny (see below). Colonies randomlyharvested from cultures of 5-AzaC-treated cells stimulatedwith either IL-3 or GM-CSF were processed in parallel ascontrol.

"-globin is a gene expressed at high levels in erythroidcells. However, low levels of expression are randomlyactivated also in very primitive hemopoietic cells [32].Therefore, it is robust levels of expression rather than"-globin expression per se that represent a marker forerythroid differentiation. RT-PCR is a very sensitivetechnique which, if not performed carefully, overesti-mates levels of gene expression. Since the amount ofcDNA obtainable from half of a single colony (%100cells) was too low for quantitative RT-PCR analysis,preliminary experiments were performed to determinethe numbers of PCR cycles necessary to maximize dif-ferences in amplification between rare and abundant"-globin cDNA. The amplification kinetics of "-globinusing, as template, cDNA extracted from cells expressinglevels of "-globin detectable (the erythroid 32D Eposubclone) or not (the nonerythroid 32D Ro cells) byNorthern analysis [11] is shown in Fig. 4B. "-actin,which was expressed at the same level by the two celllines, was amplified in parallel as controls. While "-actinwas amplified with the same kinetics with both cDNAtemplates, the kinetics of amplification of "-globin frag-ments using these same cDNAs was very different. UsingcDNA from 32D Ro, the reaction had an exponentialamplification kinetics that reached plateau after 30–35cycles. In contrast, using cDNA from 32D EPO cells, theamplification was in plateau already after 15 cycles. Onthe basis of these data, 20 amplification cycles werechosen to compare the levels of "-globin expressed byclones of 5-AzaC-treated 32D Ro cells selected with IL-3or EPO.Maximal levels of "-globin fragments were amplified

after just 20 cycles using cDNA prepared from all of thecolonies growing in EPO which have been analyzed (Fig.4A). Some levels of amplification were also obtained withcDNA extracted from colonies selected with IL-3 but theselevels were never higher than those observed with the pa-rental untreated 32D Ro cells (Fig. 4A). Therefore, all of thecolonies selected with EPO from 5-AzaC-treated 32D Rocells, but none of the colonies selected with IL-3, expressedhigh levels of "-globin.

Fig. 4. (A) RT-PCR analysis for the expression of "-globin in fivesingle IL-3-dependent (a1–a5) or EPO-dependent (b1–b5) coloniesderived from 5-AzaC-treated (3 #M for 2 days) 32D Ro cells. Ampli-fication with cDNA from the erythroid 32D Epo cells and fromuntreated 32D Ro cells is presented for comparison. cDNA wasamplified for 20 cycles and the specificity of the amplified bandwas confirmed by hybridization with an oligonucleotide spanning acDNA region internal to the primers used for the amplification(not shown). Similar results were obtained in two additional experi-ments for a total of 15 single EPO-dependent and 15 IL-3-dependentcolonies analyzed. Size markers are presented on the left. (B) Ampli-fication kinetics for "-globin (top panel) and "-actin (bottompanel) fragments using as template cDNA from erythroid (32D Epo)and nonerythroid (32D Ro) cells. The expression levels detectedwith the two cell lines by Northern are presented in the respectiveinserts for comparison.

263M. Baiocchi et al. / Experimental Cell Research 285 (2003) 258–267

EPO-dependent, but not GM-CSF- or IL-3-dependent, celllines obtained from 5-AzaC-treated 32D Ro cells areerythroid

All of EPO-dependent colonies obtained from 5-AzaC-treated 32D Ro cells generated cell lines strictly dependenton the presence of EPO for proliferation and survival whentransferred in liquid culture (none of the 10 individualclones analyzed survived for more than 48 h in the absenceof EPO or was capable of forming colonies in response toGM-CSF or G-CSF, data not shown). Their growth curveand their EPO sensitivity were absolutely identical to thosealready published for the EPO-dependent subclones derivedfrom 32D [11] and, therefore, are not shown. Their pheno-type was characterized both at the molecular and morpho-logical level using cell lines derived from colonies whichhad been concurrently selected with either IL-3 or GM-CSFas control. The molecular characterization was representedby gene expression analysis after 15 days, by RT-PCR (Fig.5), or after 1–2 months, by Northern analysis (Fig. 6), ofregular passages in the presence of EPO. The morphologicalcharacterization included May–Grunwald and benzidinestaining of cytocentrifuged smears.High levels of !- and "-globin, but very low levels of

myeloperoxidase, were detected by RT-PCR (Fig. 5) andNorthern analysis (Fig. 6) with all of the cell lines whichhad been derived from colonies selected with EPO. In con-trast, cell lines derived from the same 5-AzaC-treated cellsbut selected with either IL-3 or GM-CSF expressed highlevels of myeloperoxidase but very low levels of globingenes (Figs. 5 and 6). No differences were observed byRT-PCR in the amplification of Gata1 and EpoR betweenEPO- or IL-3/GM-CSF-dependent clones (Fig. 5). How-ever, since the levels of Gata1 and EpoR fragments wereanalyzed after 30 cycles of amplification, no conclusionmay be drawn from these data on the relative expression ofthese genes in the different lines. By Northern analysis, thelevels of expression of Gata1 and EpoR in EPO-dependentclones were consistently higher than those in GM-CSF/IL-3-dependent clones or in the original 32D Ro cell line itself(Fig. 6).The erythroid nature of the EPO-dependent cell lines

derived from 5-AzaC-treated 32D Ro cells was furtherproven by morphological analysis. While the IL-3-depen-dent clones retained the mast cell morphology of the orig-inal 32D Ro cells and the GM-CSF-dependent clones ac-quired the morphology of myelo-monocytic cells, the EPO-dependent cells had a clear erythroid morphology (data notshown). Interestingly, all of the stages of erythroid matura-tion (have pro-basophilic, polycromatophil, ortocromaticcells down to the erythroblast stage) were identified indi-cating that these cells actually undergo maturation in vitroin the presence of EPO. Conversely, benzidine-positivecells were consistently detected in EPO-dependent cell lines(10–20%) but were never present among the IL-3- or GM-CSF-dependent cells.

Discussion

It has been recently demonstrated that adult marrow,brain, and skeletal muscle contain cells capable of givingrise in vivo and in vitro to mature cellular progeny ofdisparate embryologic derivations. As an example, marrowstem cells are capable of differentiation not only into he-matopoietic cells but also into cells of the microglia andmacroglia of the brain [33], myocytes of the skeletal muscle[34] and of the myocardium [35], and as hepatocytes [36].These results prompt questioning of the unidirectionality ofthe differentiation process and development of the conceptsof “cell plasticity” and “cell transdifferentiation” [37]. Themajor skepticism toward these concepts derives from thefact that, despite the numerous experiments designed toidentify how lineage-specific differentiation is achieved, themolecular mechanism of lineage specification remains elu-sive. The observation that murine embryonic fibroblastsacquire the potential to differentiate into different mesen-

Fig. 5. RT-PCR of the expression of erythroid ("- and !-globin, Gata1,EpoR) and myelo-monocytic (myeloperoxidase) markers in three separatecell lines derived from 5-AzaC-treated 32D Ro cells selected with EPO,IL-3, or GM-CSF, as indicated. Actin was amplified as control of thecDNA quality. The analysis was performed 10 days after the cells har-vested from the single colonies had been transferred into liquid cultures.rRNA and cDNA from untreated 32D Ro cells or from 32D Epo cells werealso analyzed, as negative and positive controls, respectively. "- and!-globin fragments were amplified for 20 cycles while all of the othergenes were amplified for 30 cycles. All the fragments were hybridized witha radioactive internal oligonucleotide probe with the exception of actinwhich was stained with ethidium bromide. Similar results were obtained intwo additional experiments.

264 M. Baiocchi et al. / Experimental Cell Research 285 (2003) 258–267

chymal cell types, including myoblasts, chondrocytes, andadipocytes, following reactivation of MyoD [16], had orig-inally championed the hypothesis that lineage-specific dif-ferentiation might be achieved through activation of a dom-inant “master regulatory” gene.Gata-1, a member of a highly conserved family of tran-

scription factors [4], was originally cloned as an erythroidmaster regulatory gene [25]. In fact, Gata1 cognate se-quences are present in the regulatory regions of all of theerythroid genes identified up to date, including EpoR[38,39] and Gata1 itself [40,41]. Furthermore, mice whoseexpression of either Gata1 [42] or its obligatory partner Fog[43] has been impaired by gene disruption die prenatallydue to severe anemia. However, expression of Gatal is firstactivated at the level of the common progenitor cell [3] andis maintained in cells maturing toward all of the myeloidlineages [44]. Furthermore, additional gene disruption stud-ies have shown that a defective Gatal expression hampersthe differentiation of not only erythroid cells but also that ofmegakaryocytes [45], eosinophils [46], and mast cells [47].However, while in the majority of these lineages Gata1expression decreases with maturation, in the case of theerythroid lineage its expression increases with progressiontoward differentiation [48,49]. This observation promptedinvestigation of the effects of the levels of Gata1 expression

on lineage commitment. Avian myelo-monocytic cell linestransfected with Gata1 differentiated into hemopoietic cellsthe phenotype of which was linked to the levels of Gata1ectopic expression with only those cells expressing thehighest levels being erythroid [50]. Similarly, sublethallyirradiated mice engrafted with stem cells transfected with aGata1-containing retrovirus expressed white blood cellcounts lower than and red blood cell counts higher thanthose of animals reconstituted with normal stem cells [51].On the basis of these experiments, it has been proposed thatit is the relative concentration of a transcription factor withrespect to that of a few key partners, rather than expressionper se, that leads to lineage specification [52]. It is not clear,however, what would be a physiologic mechanism thatwould alter the relative concentration of transcription factorto induce transdifferentiation. In fact, all the mechanismsused in the experiments described above are artificial andunlikely to occur spontaneously in vivo. To further compli-cate the question, any mechanism involved in transdiffer-entiation of normal cells must take also into account the roleexerted by lineage-specific growth factors [53].To identify mechanisms that might lead to natural cellu-

lar transdifferentiation, we have exploited a lineage restric-tion that had occurred spontaneously when 32D cells werecultured for a long time with G-CSF (Ref. 12 and Fig. 2A).We assumed that G-CSF treatment might have turned themultipotent 32D cell line into myeloid-restricted 32D Rocells by inducing methylation of key regulatory gene(s). Toprove this hypothesis, we have adopted a strategy similar tothat used to differentiate fibroblastic into myoblastic cells[15], i.e., short-term incubation with the demethylatingagent 5-AzaC. In preliminary experiments, it was excludedthat the loss of erythroid differentiation potential of 32Dcells was not simply due to lack of expression of erythroidtranscription factors. These experiments demonstrated thatGata2, Scl, Nfe2, and Id were expressed at comparablelevels by 32D and 32D Ro cells (Fig. 2B). On the otherhand, comparable low levels of Gata1 were expressed byboth lines while the erythroid transcription factor Eklf wasexpressed mainly by 32D Ro cells and !-globin was ex-pressed only by the original 32D cells. Since during ery-throid differentiation activation of !-globin precedes that of"-globin [54], this last result conferms at the molecularlevel that 32D, but not 32D Ro, cells have erythroid poten-tial.Short-term incubation with concentrations of 5-AzaC

similar to those used in [15] transdifferentiated 32D Ro cellsfrom myeloid-restricted to multipotent progenitor cells. Infact, EPO-responsive clones were consistently isolated from5-AzaC-treated 32D Ro cells (Fig. 3). These clones ex-pressed high levels of "-globin shortly after the selection (1week) (Fig. 4A) and consistently gave rise to EPO-depen-dent cell lines that had the morphology of benzidine-posi-tive erythroblasts, at all stages of maturation (data notshown); expressed high levels of erythroid genes (!- and"-globin, Gata1, and EpoR); and did not express myeloper-

Fig. 6. Northern analysis of the expression of erythroid ("- and !-globin,Gata1, and EpoR) and myelo-monocytic (myeloperoxidase) markers inindependently isolated subclones of 5-AzaC-treated 32D Ro cells isolatedwith GM-CSF (4 clones), IL-3 (4 clones), or EPO (one clone at twodifferent time points in culture). The expression of the same genes in5-AzaC-treated 32D Ro and in 32D Epo cells is also presented for com-parison. The ethidium bromide staining of the gel before blotting is alsoshown as control of the total RNA loaded in each lane. Similar results wereobtained in two additional experiments.

265M. Baiocchi et al. / Experimental Cell Research 285 (2003) 258–267

oxidase (Figs. 5 and 6). Interesting, 5-AzaC-treated 32D Rocells did not loose their capacity to respond to IL-3 andGM-CSF and gave rise also to IL-3- or GM-CSF-dependentclones that were negative or expressed low levels of globingenes and EpoR (Figs. 3–6). Furthermore, all of the GM-CSF-selected clones expressed high levels of myeloperoxi-dase (Figs. 5 and 6) and had the morphology of myelo-monocytic cells (data not shown). The clones selected withIL-3 retained the mast cell phenotype of the original cellsand only one (out of more than 10 independently isolatedclones) expressed myeloperoxidase by Northern (Fig. 6).Taken together, these data suggests that G-CSF repressedthe potential of 32D cells to activate the erythroid programvia a methylation-dependent mechanism that was revertedby 5-AzaC treatment. We believe that such a reversionoccurred though a “physiologic” pathway because thetreated cells expressed the erythroid program only whenselected with EPO. There is not a consensus on the nature ofthe association between response to EPO and erythroidphenotype. Previous data [55] had demonstrated that, al-though the myeloid-restricted 32D subclones express EpoRmRNA and protein, only its erythroid subclones translocatethe receptor on the cell surface. We had, therefore, sug-gested the existence of an erythroid-specific EpoR chaper-one whose expression could represent the link betweendifferentiation program and cell potential to respond to EPO[55]. Alternatively, it has been suggested as the link amechanism that would couple the differentiation programwith the EPO signal transduction pathway as main controlfor survival and proliferation [56,57]. The fact that theEPO-dependent clones expressed maximal levels of ery-throid markers shortly after the selection is consistent withboth hypotheses.In view of the in vitro transdifferentiation experiments

described earlier and the high levels of Gata1 expressionconsistently found in Epo-responsive clones isolated from5-AzaC-treated 32D Ro cells, it is very tempting to suggestthat G-CSF treatment induced methylation of the Gata1gene itself. This hypothesis leads to an interesting explana-tion for the still poorly understood phenomenon of growthfactor priming by establishing a new link between lineage-specific growth factors and transcription factors. Unfortu-nately, although the regulatory sequences of Gata1 are wellknown [4], its functional methylation sites have been poorlystudied up to now and further experiments are necessary toclarify this point.In conclusion, we provide data suggesting a mechanism

for the G-CSF restriction of the differentiation potential in32D cells which involves methylation of gene(s) importantfor erythroid differentiation, possibly Gata1.

Acknowledgments

This study was supported by Progetti di Ricerca di In-teresse Nazionale 1999, 2000, and 2001 and Ricerca Cor-

rente 2000–2001, Ministero Salute, and Progetto CNR-MIUR SP4 “Oncologia”, Ministero Istruzione, Universita eRicerca, Rome, Italy.

References

[1] E.A. McCulloch, Stem cells in normal and leukemic hemopoiesis,Blood 62 (1983) 1–13.

[2] M. Ogawa, Differentiation and proliferation of hematopoietic stemcells, Blood 81 (1993) 2844–2853.

[3] K. Akashi, D. Traver, T. Miyamoto, I.L. Weissman, A clonogeniccommon myeloid progenitor that gives rise to all myeloid lineages,Nature 404 (2000) 193–197.

[4] S.H. Orkin, Transcription factors that regulate lineage decision, in: G.Stamatoyannopoulos, P.W. Majerus, R. Perlmutter, H. Varmus(Eds.), The Molecular Basis of Blood Diseases, Saunders, Philadel-phia, 2001.

[5] T. Enver, M. Greaves, Loops, lineage, and leukemia, Cell 94 (1998)9–12.

[6] R. Singal, G.D. Ginder, DNA methylation, Blood 93 (1999) 4059–4070.

[7] G. Migliaccio, A.R. Migliaccio, J.W. Adamson, The biology of he-matopoietic growth factors: studies in vitro under serum-deprivedconditions, Exp. Hematol 18 (1990) 1049–1055.

[8] R.L. Phillips, R.E. Ernst, B. Brunk, N. Ivanova, M.A. Mahan, J.K.Deanehan, K.A. Moore, G.C. Overton, I.R. Lemischka, The geneticprogram of hematopoietic stem cells, Science 288 (2000) 1635–1640.

[9] J.S. Greenberger, M.A. Sakakeeny, R.K. Humphries, C.J. Eaves, R.J.Eckner, Demonstration of permanent factor-dependent multipotential(erythroid/neutrophil/basophil) hematopoietic progenitor cell lines,Proc. Natl. Acad. Sci. USA 80 (1983) 2931–2935.

[10] T.M. Dexter, J. Garland, D. Scott, E. Scolnick, D. Metcalf, Growth offactor-dependent hemopoietic precursor cell lines, J. Exp. Med. 152(1980) 1036–1047.

[11] G. Migliaccio, A.R. Migliaccio, B.L. Kreider, G. Rovera, J.W. Ad-amson, Selection of lineage-restricted cell lines immortalized at dif-ferent stages of hematopoietic differentiation from the murine cellline 32D, J. Cell Biol. 109 (1989) 833–841.

[12] M. Valtieri, D.J. Tweardy, D. Caracciolo, K. Johnson, F. Mavilio, S.Altmann, D. Santoli, G. Rovera, Cytokine-dependent granulocyticdifferentiation. Regulation of proliferative and differentiative re-sponses in a murine progenitor cell line, J. Immunol. 138 (1987)3829–3835.

[13] B.L. Kreider, P.D. Phillips, M.B. Prystowsky, N. Shirsat, J.H. Pierce,R. Tushinski, G. Rovera, Induction of the granulocyte–macrophagecolony-stimulating factor (CSF) receptor by granulocyte CSF in-creases the differentiative options of a murine hematopoietic progen-itor cell, Mol. Cell Biol. 10 (1990) 4846–4853.

[14] B.L. Kreider, R. Benezra, G. Rovera, T. Kadesch, Inhibition ofmyeloid differentiation by the helix-loop-helix protein Id, Science255 (1992) 1700–1702.

[15] S.M. Taylor, P.A. Jones, Multiple new phenotypes induced in 10T1/2and 3T3 cells treated with 5-azacytidine, Cell 17 (1979) 771–779.

[16] P.A. Jones, M.J. Wolkowicz, W.M. Rideout, F.A. Gonzales, C.M.Marziasz, G.A. Coetzee, S.J. Tapscott, De novo methylation of theMyoD1 CpG island during the establishment of immortal cell lines,Proc. Natl. Acad. Sci. USA 87 (1990) 6117–6121.

[17] G. Migliaccio, A.R. Migliaccio, K. Kaushansky, J.W. Adamson,Regulation of differentiation of murine progenitor cells derived fromblast cell colonies under serum-deprived conditions, Exp. Hematol.17 (1989) 110–115.

[18] G. Migliaccio, A.R. Migliaccio, J.W. Visser, Synergism betweenerythropoietin and interleukin-3 in the induction of hematopoieticstem cell proliferation and erythroid burst colony formation, Blood 72(1988) 944–951.

266 M. Baiocchi et al. / Experimental Cell Research 285 (2003) 258–267

[19] P. Chomcynski, N. Sacchi, Single-step method of RNA isolated byacid guanidinium thiocyanate-phenol-chloroform extraction, Annu.Rev. Biochem. 162 (1987) 156–159.

[20] E. Ashihara, A.M. Vannucchi, G. Migliaccio, A.R. Migliaccio,Growth factor receptor expression during in vitro differentiation ofpartially purified populations containing murine stem cells, J. CellPhysiol. 171 (1997) 343–356.

[21] A.M. Vannucchi, F. Paoletti, S. Linari, C. Cellai, R. Caporale, P.R.Ferrini, M. Sanchez, G. Migliaccio, A.R. Migliaccio, Identificationand characterization of a bipotent (erythroid and megakaryocytic) cellprecursor from the spleen of phenylhydrazine-treated mice, Blood 95(2000) 2559–2568.

[22] J. Sambrook, E.F. Fritsch, T. Maniatis, Molecular Cloning: A Labo-ratory Manual, 2nd ed, Cold Spring Harbor Laboratory Press, ColdSpring Harbor, New York, 1989.

[23] C.G. Begley, J. Visvader, A.R. Green, P.D. Aplan, D. Metcalf, I.R.Kirsch, N.M. Gough, Molecular cloning and chromosomal localiza-tion of the murine homolog of the human helix-loop-helix gene SCL,Proc. Natl. Acad. Sci. USA 88 (1991) 869–873.

[24] R. Benezra, R.L. Davis, D. Lockshon, D.L. Turner, H. Weintraub,The protein Id: a negative regulator of helix–loop–helix DNA bind-ing proteins, Cell 61 (1990) 49–59.

[25] S.F. Tsai, D.I. Martin, L.I. Zon, A.D. D’Andrea, G.G. Wong, S.H.Orkin, Cloning of cDNA for the major DNA-binding protein of theerythroid lineage through expression in mammalian cells, Nature 339(1989) 446–451.

[26] F.Y. Tsai, G. Keller, F.C. Kuo, M. Weiss, J. Chen, M. Rosenblatt,F.W. Alt, S.H. Orkin, An early haematopoietic defect in mice lackingthe transcription factor GATA-2, Nature 371 (1994) 221–226.

[27] I.J. Miller, J.J. Bieker, A novel, erythroid cell-specific murine transcrip-tion factor that binds to the CACCC element and is related to the Kruppelfamily of nuclear proteins, Mol. Cell Biol. 13 (1993) 2776–2786.

[28] J.Y. Chan, X.L. Han, Y.W. Kan, Isolation of cDNA encoding the humanNF-E2 protein, Proc. Natl. Acad. Sci. USA 90 (1993) 11366–11370.

[29] A.D. D’Andrea, H.F. Lodish, G.G. Wong, Expression cloning of themurine erythropoietin receptor, Cell 57 (1989) 277–285.

[30] R.G. Ramsay, K. Ikeda, R.A. Rifkind, P.A. Marks, Changes in geneexpression associated with induced differentiation of erythroleuke-mia: protooncogenes, globin genes, and cell division, Proc. Natl.Acad. Sci. USA 83 (1986) 6849–6853.

[31] S.H. Orkin, F.I. Harosi, P. Leder, Differentiation in erythroleukemic cellsand their somatic hybrids, Proc. Natl. Acad. Sci. USA 72 (1975) 98–102.

[32] M. Hu, D. Krause, M. Greaves, S. Sharkis, M. Dexter, C. Heyworth,T. Enver, Multilineage gene expression precedes commitment in thehemopoietic system, Genes Dev. 11 (1997) 774–785.

[33] M.A. Eglitis, E. Mezey, Hematopoietic cells differentiate into bothmicroglia and macroglia in the brains of adult mice, Proc. Natl. Acad.Sci. USA 94 (1997) 4080–4085.

[34] G. Ferrari, A.G. Cusella-De, M. Coletta, E. Paolucci, A. Stornaiuolo,G. Cossu, F. Mavilio, Muscle regeneration by bone marrow-derivedmyogenic progenitors, Science 279 (1998) 1528–1530.

[35] D. Orlic, J. Kajstura, S. Chimenti, I. Jakoniuk, S.M. Anderson, B. Li,J. Pickel, R. McKay, B. Nadal-Ginard, D.M. Bodine, A. Leri, P.Anversa, Bone marrow cells regenerate infarcted myocardium, Na-ture 410 (2001) 701–705.

[36] E. Lagasse, H. Connors, M. Al-Dhalimy, M. Reitsma, M. Dohse, L.Osborne, X. Wang, M. Finegold, I.L. Weissman, M. Grompe, Puri-fied hematopoietic stem cells can differentiate into hepatocytes invivo, Nat. Med. 6 (2000) 1229–1234.

[37] G. Almeida-Porada, C. Porada, E.D. Zanjani, Adult stem cell plasticityand methods of detection, Rev. Clin. Exp. Hematol. 5 (2001) 26–41.

[38] L.I. Zon, H. Youssoufian, C. Mather, H.F. Lodish, S.H. Orkin, Ac-tivation of the erythropoietin receptor promoter by transcription fac-tor GATA-1, Proc. Natl. Acad. Sci. USA 88 (1991) 10638–10641.

[39] T. Chiba, Y. Ikawa, K. Todokoro, GATA-1 transactivates erythro-poietin receptor gene, and erythropoietin receptor-mediated signals

enhance GATA-1 gene expression, Nucleic Acids Res. 19 (1991)3843–3848.

[40] S.F. Tsai, E. Strauss, S.H. Orkin, Functional analysis and in vivofootprinting implicate the erythroid transcription factor GATA-1 as apositive regulator of its own promoter, Genes Dev. 5 (1991) 919–931.

[41] M.H. Baron, S.M. Farrington, Positive regulators of the lineage-specific transcription factor GATA-1 in differentiating erythroid cells,Mol. Cell Biol. 14 (1994) 3108–3114.

[42] L. Pevny, M.C. Simon, E. Robertson, W.H. Klein, S.F. Tsai, V.D’Agati, S.H. Orkin, F. Costantini, Erythroid differentiation in chi-maeric mice blocked by a targeted mutation in the gene for transcrip-tion factor GATA-1, Nature 349 (1991) 257–260.

[43] A.P. Tsang, Y. Fujiwara, D.B. Hom, S.H. Orkin, Failure of megakaryo-poiesis and arrested erythropoiesis in mice lacking the GATA-1 tran-scriptional cofactor FOG, Genes Dev. 12 (1998) 1176–1188.

[44] S. Crotta, S. Nicolis, A. Ronchi, S. Ottolenghi, L. Ruzzi, Y. Shimada,A.R. Migliaccio, G. Migliaccio, Progressive inactivation of the expres-sion of an erythroid transcriptional factor in GM- and G-CSF-dependentmyeloid cell lines, Nucleic Acids Res. 18 (1990) 6863–6869.

[45] R.A. Shivdasani, Y. Fujiwara, M.A. McDevitt, S.H. Orkin, A lineage-selective knockout establishes the critical role of transcription factorGATA-1 in megakaryocyte growth and platelet development, EMBOJ. 16 (1997) 3965–3973.

[46] C. Yu, A.B. Cantor, H. Yang, C. Browne, R.A. Wells, Y. Fujiwara,S.H. Orkin, Targeted deletion of a high-affinity GATA-binding site inthe GATA-1 promoter leads to selective loss of the eosinophil lineagein vivo, J. Exp. Med. 195 (2002) 1387–1395.

[47] Migliaccio, A.R. Rana, R.A. Sanchez, M. Lorenzini, R. Centurione,L. Bianchi, L., Vannucchi, A.M. Migliaccio, G. GATA-1 as a regu-lator of mast cell differentiation revealed by the phenotype of theGATA-low mouse mutant, J. Exp. Med. 197 (2003) 281–296.

[48] M.A. McDevitt, R.A. Shivdasani, Y. Fujiwara, H. Yang, S.H. Orkin,A “knockdown” mutation created by cis-element gene targeting re-veals the dependence of erythroid cell maturation on the level oftranscription factor GATA-1, Proc. Natl. Acad. Sci. USA 94 (1997)6781–6785.

[49] A. Vannucchi, S. Linari, C-S. Lin, M. Koury, M. Bondurant, A.R.Migliaccio, Increased expression of the distal, but not of the proximalGata1 transcripts during differentiation of erythroid cells, J. CellPhys. 180 (1999) 390–401.

[50] H. Kulessa, J. Frampton, T. Graf, GATA-1 reprograms avian my-elomonocytic cell lines into eosinophils, thromboblasts, and erythro-blasts, Genes Dev. 9 (1995) 1250–1262.

[51] S.F. Farina, L.J. Girard, E.F. Vanin, A.W. Nienhuis, D.M. Bodine,Dysregulated expression of GATA-1 following retrovirus-mediatedgene transfer into murine hematopoietic stem cells increases eryth-ropoiesis, Blood 86 (1995) 4124–4133.

[52] K. McNagny, T. Graf, Making eosinophils through subtle shifts intranscription factor expression, J. Exp. Med. 195 (2002) F43–F47.

[53] D. Metcalf, T. Enver, C.M. Heyworth, T.M. Dexter, Lineage com-mitment and maturation in hematopoietic cells. The case for extrinsicregulation. Do stem cells play dice? Blood 92 (1998) 345–348.

[54] N. Beru, P.B. Maples, O. Hermine, E. Goldwasser, Differential ex-pression of alpha- and beta-globin genes in erythroleukemic cell lines,Mol. Cell Biol. 10 (1990) 3591–3595.

[55] A.R. Migliaccio, G. Migliaccio, A. D’Andrea, M. Baiocchi, A. Ron-chi, S. Nicolis, S. Ottolenghi, J.W. Adamson, Response to erythro-poietin in erythroid subclones of the factor-dependent cell line 32D isdetermined by translocation of the erythropoietin receptor to the cellsurface, Proc. Natl. Acad. Sci. USA 88 (1991) 11086–11090.

[56] M.J. Koury, M.C. Bondurant, Erythropoietin retards DNA breakdownand prevents programmed death in erythroid progenitor cells, Science248 (1990) 378–381.

[57] J.L. Spivak, T. Pham, M. Isaacs, W.D. Hankins, Erythropoietin isboth a mitogen and a survival factor, Blood 77 (1991) 1228–1233.

267M. Baiocchi et al. / Experimental Cell Research 285 (2003) 258–267