Prospective isolation of human erythroidlineage-committed progenitorsYasuo Moria,b, James Y. Chena,b, John V. Pluvinagea,b, Jun Seitaa,b,1,2, and Irving L. Weissmana,b,1,2

aInstitute for Stem Cell Biology and Regenerative Medicine, Stanford University School of Medicine, Stanford, CA 94305; and bLudwig Center for CancerStem Cell Biology and Medicine at Stanford University, Stanford, CA 94305

Contributed by Irving L. Weissman, June 21, 2015 (sent for review February 17, 2015)

Determining the developmental pathway leading to erythrocytesand being able to isolate their progenitors are crucial to un-derstanding and treating disorders of red cell imbalance such asanemia, myelodysplastic syndrome, and polycythemia vera. Herewe show that the human erythrocyte progenitor (hEP) can beprospectively isolated from adult bone marrow. We found threesubfractions that possessed different expression patterns of CD105and CD71 within the previously defined human megakaryocyte/erythrocyte progenitor (hMEP; Lineage− CD34+ CD38+ IL-3Rα−

CD45RA−) population. Both CD71− CD105− and CD71+ CD105−

MEPs, at least in vitro, still retained bipotency for the megakaryo-cyte (MegK) and erythrocyte (E) lineages, although the latter sub-population is skewed in differentiation toward the erythroidlineage. Notably, the proliferative and differentiation output of theCD71intermediate(int)/+ CD105+ subset of cells within the MEP popu-lation was completely restricted to the erythroid lineage with theloss of MegK potential. CD71+ CD105− MEPs are erythrocyte-biased MEPs (E-MEPs) and CD71int/+ CD105+ cells are EPs. Thesepreviously unclassified populations may facilitate further under-standing of the molecular mechanisms governing human erythroiddevelopment and serve as potential therapeutic targets in disor-ders of the erythroid lineage.

erythroid progenitor | endoglin | lineage commitment | hematopoiesis |transcription factor

It is now apparent that hematopoiesis derives throughoutpostnatal life from the constant input from a small fraction of

hematopoietic stem cells (HSCs) (1, 2), themselves rare, un-dergoing cell divisions that include self-renewal to HSCs anddifferentiation. The differentiation largely to multipotent pro-genitors (MPPs) that undergo transient amplifying divisions togive rise to all blood cell types but do not undergo long-term self-renewal (3). In mice and humans, the steps between HSCs andmature blood cells are stepwise quantal genetic/epigeneticchanges with bifurcations to produce progeny with more limitedfates; these are the oligopotent and unipotent progenitors, nowalmost fully described in mouse hematopoiesis (4). Determiningthe differentiation pathway leading to erythrocytes and isolatingerythroid-specific progenitors are crucial to understanding howfate determinations are made that result in homeostatic pro-duction of blood cells and elements. Such an understanding canalso elucidate disorders of red cell imbalance, such as anemia,myelodysplastic syndrome (MDS), and polycythemia vera (PV)(5, 6). Here we sought to find cell-surface markers that can beused to isolate from human bone marrow each stage of differ-entiation after the fate determinations that lead MPPs to com-mon myeloid progenitors (CMPs) to megakaryocyte/erythroidprogenitors (MEPs) and allow the commitment to erythropoiesis.Certain aspects of erythrocyte development have been elucidated,such as the overall pathway from HSCs through CMPs and MEPs(7–9). However, the original human CMP and MEP populationsstill appear to be heterogeneous, and the subsequent stages ofdifferentiation (i.e., from MEP to erythrocyte) remain unclear.Given this background, we sought markers that could sub-

divide CMPs and/or MEPs into cells specifically destined to

generate erythrocytes. Erythrocytes, like all blood lineages, de-velop through a series of differentiation stages that begin withHSCs, which have been prospectively isolated in humans (10). Thesurface markers of intermediate erythroid progenitors/precursorsand the expression pattern of some pivotal transcription factors(TFs) during their differentiation from HSCs have been analyzed(11–19). As a result, the erythroid-committed progenitor (EP),which exists downstream of the bipotent MEP, has been isolatedin mouse but not human (9).To isolate the presumed human erythrocyte progenitor (hEP),

we first looked at markers crucial to isolating the mouse EP. Onesuch key marker is CD105 (endoglin), whereby mouse EPs areLineage (Lin)− stem cell antigen (Sca)-1− receptor tyrosine kinasec-Kit+ CD16/32− CD150+ CD105+ CD41−. In humans, CD105and CD71 have been classified as early erythroid cell markers(14–19). However, CD105 and CD71 expression in oligopotentmyeloid progenitors (8) [i.e., CMPs, granulocyte/macrophage pro-genitors (GMPs), and MEPs] has not been evaluated. Thus, weexamined closely the expression of these markers in CMPs, GMPs,MEPs, and their progeny.

ResultsCD71 and CD105 Expression Patterns Subfractionate Myeloid Progenitors.First, CMPs, GMPs, and MEPs were purified from human bonemarrow (BM) according to the flow cytometric sorting schemeshown in Fig. 1A. The BM mononuclear cell fraction that isnegative for lineage-affiliated antigens (Lin−) was subdividedinto the CD34+ CD38− CD45RA− HSC/MPP and CD34+

CD38+ progenitor fractions (10, 20). The Lin− CD34+ CD38+

progenitors were then fractionated into CMP, GMP, and MEPpopulations according to the expression patterns of IL-3Rα andCD45RA, as previously reported (8): CMPs, GMPs, and MEPswere defined as IL-3Rα+ CD45RA−, IL-3Rα+ CD45RA+, andIL-3Rα−/lo CD45RA− populations, respectively.

Significance

We have identified the first step of erythrocyte lineage com-mitment in human bone marrow as a CD71intermediate/+ CD105+

cell fraction of a previously defined megakaryocyte/erythro-cyte progenitor population. This purification could be a usefultool for studying physiological and pathological red blood celldevelopment, and should be analyzed in patients sufferingfrom anemia or erythrocytosis such as in myelodysplastic syn-drome or polycythemia vera.

Author contributions: Y.M., J.Y.C., J.V.P., J.S., and I.L.W. designed research; Y.M., J.Y.C., J.V.P.,and J.S. performed research; Y.M., J.Y.C., J.V.P., and J.S. analyzed data; and Y.M., J.Y.C., J.V.P.,J.S., and I.L.W. wrote the paper.

The authors declare no conflict of interest.

Freely available online through the PNAS open access option.1J.S. and I.L.W. contributed equally to this work.2To whom correspondence may be addressed. Email: jseita@stanford.edu or irv@stanford.edu.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1512076112/-/DCSupplemental.

9638–9643 | PNAS | August 4, 2015 | vol. 112 | no. 31 www.pnas.org/cgi/doi/10.1073/pnas.1512076112

Dow

nloa

ded

by g

uest

on

Janu

ary

26, 2

022

We analyzed the expression pattern of putative erythroidmarkers for each of the progenitor populations. The expressionof CD71 was detectable in the CMP (31.5 ± 11.1%, n = 8; Fig.1A represents one of these experiments) and MEP fractions(82.3 ± 6.4%) but in very few cells in the CD34+ CD38− HSC/MPP (5.8 ± 2.8%) and GMP fractions (5.3 ± 5.4%). Amongthese populations, only MEPs expressed CD105, and CD105-positive cells coexpress CD71 at intermediate to positive in-tensity [hereafter CD71intermediate(int)/+ CD105+ MEPs; 34.5 ±6.9% of MEPs]. CD36, a known early marker of erythroid de-velopment (19, 21, 22), was expressed in some CD71+ CD105−

MEPs but almost all CD71int/+ CD105+ MEPs. In contrast,glycophorin A (GPA/CD235), a late marker of erythroid de-velopment (11), was not detectable in any fractions examined(Fig. 1B). Subfractions based on CD71 and CD105 expressionwere also found in cord blood, although the CD71int/+ CD105+

MEP fraction was much smaller in cord blood than in adultbone marrow (Fig. S1).

CD71 Expression Initiated at the CMP Stage Represents Megakaryocyte/Erythrocyte-Biased Lineage Potential. We functionally examined thedifferentiation potential of CD71− CD105− CMPs and CD71+

CD105−CMPs (Fig. 2A). Methylcellulose colony assay revealed thatcolony-forming efficiencies are decreased from CD71− CD105−

CMPs (38.2 ± 5.2%) to CD71+ CD105− CMPs (33.0 ± 2.2%)(P = 0.047) (Fig. 2B) and the differentiation potential ofCD71+ CD105− CMPs skewed toward the megakaryocyte(MegK)/erythrocyte (E) (MegE) lineage, whereas CD71− CD105−

CMPs generated a variety of myeloid colonies including colony-forming unit granulocyte/macrophage (CFU-GM) (Fig. 2 C and D).We further tested the MegK and erythroid potentials of eachfraction by a serum-free liquid culture supplemented with IL-3,stem cell factor (SCF), erythropoietin (EPO), and thrombopoietin(TPO). As shown in Fig. 2 E and F, both CD71− CD105− CMPsand CD71+ CD105− CMPs could give rise to CD41-expressingMegKs as well as GPA+ erythrocytes. However, CD71+ CD105−

CMPs gave rise to higher numbers of both CD41+ cells (P =

0.048) and GPA+ cells (P = 0.002) than did CD71− CD105−

CMPs (Fig. 2G).

Subsequent CD105 Up-Regulation Within MEPs Marks Complete ErythroidLineage Commitment. Among three subfractions of MEPs (shown inFig. 3A), the colony-forming potential was higher in the CD71+

CD105− fraction (37.0 ± 4.9%) than in the CD71int/+ CD105+

fraction (29.2 ± 4.2%) (P = 0.014) and in the CD71− CD105−

fraction (14.7 ± 3.4%) (P < 0.0001) (Fig. 2B). CD71+ CD105−MEPsgenerated primarily large E colonies [defined as burst-forming unit-erythroid colony (BFU-E)–derived], although they still retained someMegK potential (Fig. 4A). In contrast, the CD71int/+ CD105+ MEPsgenerated only the E lineage, and small-sized colonies from thisfraction contained more mature (enucleated) erythrocytes [scored ascolony-forming unit-erythroid (CFU-E)–derived] (Fig. 3 C and D).These results suggest that CD71int/+ CD105+ MEPs exist at a stagedownstream of CD71+ CD105−MEPs. A similar result was observedin a serum-free liquid culture system: CD71int/+ CD105+ MEPslost the MegK lineage potential (Fig. 3 E and F). Based on thesefindings, we designate the Lin− CD34+ CD38+ IL-3Rα−/loCD45RA− CD71+ CD105− as E-biased MEPs (E-MEPs) and theLin− CD34+ CD38+ IL-3Rα−/lo CD45RA− CD71int/+ CD105+ frac-tion as E-committed progenitors.In contrast, CD71− CD105− MEPs frequently generated MegK-

containing colonies but not CFU-GM. Serum-free liquid culturealso showed this fraction gave rise most efficiently to CD41+

MegK cells, indicating that these are not just contaminated CD71−

CD105− CMPs but might contain putative MegK progenitors.

Immunophenotypic Comparisons Among Subfractionated CMPs/MEPs.We further characterized the cell-surface markers of the CMPand MEP subfractions. Flt3/flk2 is a tyrosine kinase receptor

Lineage CD38

CMP MEP GMP

CD45RA CD45RA

CD71 CD71 CD71 CD71

A

B

CD

105

FS

C

CD

105

CD

105

CD

105

CD

34

IL-3

Rα

IL-3

Rα

Singlet PI- Lin- CD34+CD38- CD34+CD38+

CD34+CD38-

15.4%

2.95% 9.51%80.2%HSC/MPP 35.2%

CMP

GMP26.3%

MEP31.0%

96.8%

0.2%

2.5%

75.9%

1.26%

21.0%

16.4%

32.9%

46.7%

98.1%

0%

1.48%

GPACD36

CD71-CD105- CMPCD71+CD105- CMPCD71-CD105- MEPCD71+CD105- MEP

CD71int/+CD105+ MEP

FMOCD71-CD105- CMPCD71+CD105- CMPCD71-CD105- MEPCD71+CD105- MEP

CD71int/+CD105+ MEP

FMO

Erythrocyte

Fig. 1. Flow cytometric analysis of human myeloid progenitors in the bonemarrow. (A) CD71 and CD105 (endoglin) expression reveals phenotypicheterogeneity of myeloid progenitor cell compartments. (Upper) Gatingstrategy for HSCs/MPPs, CMPs, GMPs, and MEPs. (Lower) CD105 and CD71expression for each population. The percentage of events in each gate is shownin the plot. PI, propidium iodide; FSC, forward scatter. (B) Evaluation of othererythroid-affiliated surface markers CD36 and GPA. The threshold betweennegative and positive was defined by the fluorescence minus one (FMO)method, and erythrocytes were used as a positive control for GPA expression.

01020304050

020406080

100

1

10

100

1000

CD71- CD105- CD71+ CD105-

CFU-E

BFU-E± MegK

A E

G

B

FD

C

CF

Us/

100

cells

Pro

port

ion

(%)

CFU-GM

Unstainedcontrol

Stained

BFU-E

GPA

CD

41

CD71-CD105-

CD71+CD105- 0%

0%

100%

0%

0%

100%

94.7%

0.91%

3.23%

81.3%

1.74%

14.1%

CD41+

GPA+

Abs

olut

e nu

mbe

r(x

103 )

CD41+ GPA+

CD71CD105

-- -

+ CD71CD105

-- -

+

CFU-GM

CD71CD105

-- -

+ -- -

+

*

*

**

Fig. 2. CD71+ fraction within the original human CMP showed differenti-ation potential skewed toward the MegE lineage. (A and B) The morphology(A) (May–Giemsa stain) and in vitro colony-forming potential (B) of FACS-purified CD71− CD105− or CD71+ CD105− CMPs. The number of colony-forming units from 100 cells of each fraction is shown. Data presented aremean ± SD (n = 6); *P < 0.05. In this culture condition, unfractionated CMPsform 38.7 ± 4.0 colonies per 100 cells (n = 3). (Scale bars, 10 μm.) (C and D)Colonies were picked up, cytospun, and stained by the May–Giemsa methodto determine the cell types included. (E) FACS-purified 2,000 cells of eachfraction were cultured for 10 d under serum-free conditions and then ana-lyzed. CD41 or GPA positivity was compared with an unstained control (Left).(F) Cytospin preparations (May–Giemsa staining) of sorted CD41+ MegKs(Upper) and GPA+ erythrocytes (Lower) are shown. The progeny from CD71+

CD105− CMPs are shown. (G) Absolute number of CD41+ MegKs or GPA+

erythrocytes. Data shown are mean ± SD (n = 3); *P < 0.05, **P < 0.01.

Mori et al. PNAS | August 4, 2015 | vol. 112 | no. 31 | 9639

CELL

BIOLO

GY

Dow

nloa

ded

by g

uest

on

Janu

ary

26, 2

022

known for its heterogeneous expression in CMPs, high expres-sion in GMPs, and down-regulation in MegE lineages (23, 24).Among CMPs, Flt3− cells are considered to be in a transitionalstage to MEPs. In accordance with these earlier findings, our re-sults indicated that Flt3 is expressed on CD71− CMPs, thengradually down-regulated from CD71+ CMPs to MEPs but highlyexpressed on GMPs (Fig. S2A).Consistent with the in vitro MegK potential, we found that

MegK-associated molecules CD41 (13, 25) and CD9 (26–28),known to be expressed in the mouse MegK progenitor, wereexpressed at higher frequencies on the surface of human CD71+

CD105− CMPs than on CD71− CDS105− CMPs, whereas CD226and CD42b were not significantly expressed on CD71− CD105−

CMPs (∼1%) or CD71+ CD105− CMPs (∼5%). In contrast, theMegK-related markers analyzed were down-regulated or notexpressed on the surface of CD71int/+ CD105+ MEPs (Fig. S2 Band C).

Human EPs Develop from E-MEPs. To test the lineage relationshipof these subfractionated myeloerythroid progenitors directly, wereanalyzed cell-surface marker expression patterns after a short-term liquid culture. After culture for 60 h, CD71− CD105− CMPsgave rise to CD71+ CD105− CMPs, GMPs, and MEPs (bothCD71− CD105− and CD71+ CD105−), whereas CD71+ CD105−

CMPs differentiated toward MEPs but not GMPs (Fig. S3A).Furthermore, a minor fraction of progeny possessed the surfacephenotype corresponding to EPs (Fig. S3A). These data suggestedthat the CD71− CD105− CMPs are the precursor of CD71+

CD105− CMPs. However, CD71+ CD105− CMPs generated a fewCD71− CD105− CMPs, suggesting either CD71− contamination orpossible bidirectionality between them. In the same culture con-ditions, CD71+ CD105− E-MEPs differentiated mainly intoCD71int/+ CD105+ EPs and a small number of CD71− CD105−

MEPs (Fig. S3B), whereas CD71int/+ CD105+ EPs did notgenerate E-MEPs or CD71− CD105− MEPs. These data clearlyindicate that EPs exist at a stage downstream of E-MEPs. Toperform a secondary analysis, cells were sorted and subjected to

colony-forming assays (Fig. 4A). Such phenotypically defined sec-ondary myeloid progenitors displayed differentiation activity con-sistent with their original phenotypic definition. CD71-expressingcells mainly gave rise to the MegE lineage, and CD105+ cells al-most exclusively produced CFU-E–type erythroid cells.

Suppressive Effect of TGF-β on Proliferation of EPs/E-MEPs Is Accompaniedby Accelerated Terminal Differentiation. We tested the effect of TGF-β1 onMEP subfractions according to the following rationale. CD105is an accessory molecule of the transforming growth factor beta(TGF-β) type III receptor complex, able to bind TGF-β1 and TGF-β3 but not TGF-β2 (29). TGF-β, as well as tumor necrosis factoralpha and IFN gamma, is a powerful inhibitor of erythropoiesis bothin vivo and in vitro. The suppressive effect of TGF-β on E-lineagecell proliferation is accompanied by an acceleration of terminalmaturation (30–34).We found that TGF-β1 did not affect the number of colonies

generated by all three MEP subfractions (Fig. 5A) but affectedthe types of colonies generated: The proportion of CFU-E–typecolonies was increased at the expense of BFU-E–type colonies(Fig. 5 B and C). By using a serum-free suspension culture, wealso monitored the effects of TGF-β1 at an earlier time point.Additional TGF-β1 yielded two- to threefold fewer cells after 4d of culture (Fig. S4A), with higher expression levels of GPA(Fig. 5D) and morphological signs of maturation (Fig. 5E) in allsubfractions. These findings suggested that TGF-β1 provided asignal that inhibited proliferation and promoted differentiationamong CD71int/+ CD105+ EPs as well as in CD105-expressingprogenies of CD71− CD105− or CD71+ CD105− MEPs (30–34).On day 8 of the same culture, we analyzed the effect of TGF-β

on MegK development (Fig. S4B). TGF-β negatively affected theproduction of CD41+ MegKs from CD71− CD105− or CD71+

CD105− MEPs (50–65% reduction compared with controls; Fig.S4C). As a result, the frequency of CD41+ MegKs in the progenyof CD71− CD105− and CD71+ CD105− MEPs was slightly ele-vated. Adding TGF-β did not change the E lineage–restrictedpotential of EPs.

Gene Expression Analysis of Human Myeloerythroid Progenitors. Wepreviously reported that various TFs play a pivotal role in lineagespecification in both mouse (7, 35) and human (8, 36) hemato-poiesis. Each purified population described above (CMP andMEP subpopulations and GMPs) was subjected to real-timePCR analysis to test the expression profiles of TFs and lineage-related cytokine receptor genes (shown in Fig. S5). GATA familyTFs (GATA-1, GATA-2) and their cofactor Friend of GATA(FOG)-1 are required for MegE lineage development: Knockoutof GATA-1 (37, 38) or FOG-1 (39) genes is embryonically lethalin mice due to severe anemia. These GATAs were up-regulatedaccording to MegE lineage differentiation but down-regulated in

01020304050

020406080

100

110

100

1000

EA

F

CD71-CD105-

CD71+CD105-

CD71int/+CD105+

MegK ± E

CFU-E

BFU-E

D

B

MegK

Unstainedcontrol

Stained

BFU-ECFU-E

GPA

CD

41

CD71-CD105-

CD71+CD105-

CD71int/+CD105+ 0%

0%

100%

0%

0%

100%

0%

0%

100%

94.9%

1.9%

2.01%

98.5%

0.29%

0.35%

99.4%

0.01%

0.29%

Abs

olut

e nu

mbe

r (x

103 )

CD41+ GPA+

C

CF

Us/

100

cells

Pro

port

ion(

%)

CD71CD105

-- -

+ int/++

CD71CD105

-- -

+ int/++

CD71CD105

-- -

+ int/++

-- -

+ int/++

** *

Fig. 3. CD71int/+ CD105+ fraction within the original human MEP representsthe human EP. (A and B) The morphology (A) (May–Giemsa stain) andin vitro colony-forming potential (B) of FACS-purified MEP subfractions. Thenumber of CFUs from 100 cells of each fraction. Data shown are mean ± SD(n = 6); *P < 0.05, **P < 0.01. (Scale bars, 10 μm.) (C and D) Colonies werepicked up, cytospun, and stained by the May–Giemsa method to determinethe cell types included. (E) Representative flow cytometry plot of 10-d progenyof MEP subfractions. The percentage of events in each gate is shown.(F) Absolute numbers of CD41+ MegKs and GPA+ erythrocytes. Data shownare mean ± SD (n = 3).

CD71- CD105- CMPCD71+ CD105- CMP

CD71+ CD105- MEP

CD71int/+ CD105+ MEP CD71int/+ CD105+ MEP

CD71int/+ CD105+ MEP

CFUs/100 cells200 10 30 40

CD71- CD105- CMP

CD71+ CD105- CMP

CD71+ CD105- MEP

CD71int/+ CD105+ CMP/MEPCD71+ CD105- CMP/MEP

CFU-GMMegK-containingCFU-E

BFU-EPrimary Sort Secondary Sort

Fig. 4. Lineage relationship of subfractionated human myeloid progenitors.Lineage potential of FACS-purified progenitors from the culture of eachprimary fraction. After 60 h in liquid culture, cells were subfractionatedby secondary sort and subjected to colony assay. CD71int/+ CD105+ MEPsformed only CFU-E.

9640 | www.pnas.org/cgi/doi/10.1073/pnas.1512076112 Mori et al.

Dow

nloa

ded

by g

uest

on

Janu

ary

26, 2

022

the GMP stage. On the other hand, GM-affiliated TFs (i.e.,C/EBPα, PU.1) were elevated along with GM commitment fromCMPs to GMPs but suppressed in MEPs, consistent with previousreports (7, 8). Erythroid Krüppel-like factor (EKLF; KLF-1) is acritical TF for erythroid development (9, 40, 41); KLF-1 and acritical MegK activator, Friend leukemia integration (Fli)-1 (42),antagonize each other at the bifurcation of MegK versus erythroidlineages (40, 43). Consistent with their hematopoietic outcome,CD71int/+ CD105+ EPs showed the highest KLF-1 but the lowestFli-1 expression among the MEP subfractions (Fig. S5). Con-versely, CD71− CD105− MEPs showed the lowest KLF-1 but thehighest Fli-1 expression among the MEP subfractions, presumablyreflecting its robust MegK potential.EPO signaling is essential for erythroid cell development, es-

pecially following the progenitor stage, where EPO mainly blocksapoptosis (44, 45). We found that EPOR, a gene encoding theEPO receptor, was expressed at divergent levels among sub-populations: low in CMPs, higher in bipotent MegEs, and higherstill in CD71int/+ CD105+ EPs (Fig. S5). TPO is the most potentcytokine that physiologically regulates MegK and subsequentplatelet production (46). Its receptor (TPOR; c-Mpl) expressionwas lowest in CD71int/+ CD105+ EPs among the various MEPsubfractions.These findings suggest that changes in expression patterns of

lineage-instructive TFs or lineage-related cytokine receptors inMEPs are similar between human and mouse (9), although theexact identity of human MegK progenitors remains unclear.

DiscussionWe report in the present study that hEPs are prospectively iso-latable in adult steady-state bone marrow as a subset of cellsamong the originally defined hMEPs, showing that the con-ventional hMEP population is heterogeneous. Moreover, truly

bipotent hMEPs reside only in the CD105− fraction of conven-tional hMEPs (Fig. 6).CD71, the transferrin receptor, has been well-established as

one of the early E-lineage markers; the CD34+ CD71+ GPA−

fraction, defined as unipotent E-lineage progenitors/precursors,exists at a stage downstream of MEPs, and was used for functionaland/or gene expression analysis (11, 13, 30, 47). However, wefound CD71 expression at a much earlier stage of differentiationin a subset of CMPs. A fraction (∼30%) of CMPs is CD71-positivebut still retains both GM potential and MegK potential (Fig. 2).In addition, even CD71+ CD105− MEPs (E-MEPs) show MegKpotential; thus, CD71 alone may be useful for enrichment but isinsufficient for the purification of E-lineage progenitors.Combining CD71 with CD105 (endoglin) staining enables

us to identify hEPs. Most colonies that originate from CD71int/+

CD105+ hEPs are of the CFU-E type (containing a low numberof fully matured erythrocytes; Fig. 3), indicating that this isolatedfraction corresponds to mouse pre–CFU-Es (9). Importantly, hEPsnever generated MegK cells even under serum-free culture condi-tions, which allow optimal growth and differentiation of MegKprogenitors (48). CD105 is part of the TGF-β receptor complex(29); several studies have revealed an inhibitory effect of TGF-β onE-lineage cell proliferation, causing accelerated terminal matura-tion by blocking the cell cycle of immature cells (30–34). However,these studies targeted CD34+ CD71+ or CD36+ “erythroid pro-genitors” and not CD105+ cells. We found a similar effect of TGF-βon the more specific CD71int/+ CD105+ EPs (Fig. 5). Therefore, theeffect of TGF-β observed in previous studies may be largely due tothis subset, because CD71int/+ CD105+ EPs expressed both CD36and CD71 on the cell surface (shown in Fig. 1B). TGF-β is com-monly present in FCS and is known to impose negative effects notonly on erythroid but also on MegK production (48). In addition,TGF-β showed a less inhibitory effect on MegK production thanthat on erythrocyte production (Fig. 5), and thus it may reflect theabsence of CD105 expression in putative MegK progenitors.We sought to trace the developmental pathway in human

CMPs and MEPs in vitro (Fig. 4). After a 60-h culture, CD71−

CD105− CMPs could differentiate into CD71+ CD105− CMPs aswell as MEPs/GMPs. CD71+ CD105− CMPs generated more cellscorresponding to E-MEPs and EPs in addition to some CD71−

CD105−CMPs/MEPs at the same time point. Furthermore, CD71+

10

0

20

30

40

50

0

20

40

60

80

100A

D

B

C

E

GPA

CD71-CD105-

CD71int/+CD105+

CF

Us/

100

cells

Pro

port

ion(

%)

- + - + - +TGF-β

MegK ± E CFU-E BFU-E

TGF-β -

TGF-β +

CD71-CD105-

CD71int/+CD105+

TGF-β -

TGF-β +

CD71-CD105-

CD71+CD105-

CD71int/+CD105+

TGF-β - TGF-β +unstain control

CD71CD105

-- -

+int/+

+- + - + - +TGF-β

CD71CD105

-- -

+int/+

+

ns

ns

ns

Fig. 5. TGF-β1 accelerates erythroid cell maturation from all subfractions ofhuman MEPs. (A) FACS-purified MEP subfractions were cultured in methyl-cellulose-based medium for 10–12 d in the presence of a cytokine mixturewith or without TGF-β1 (2 ng/mL). The number of colony-forming units from100 cells of each fraction was assayed. The data shown are mean ± SD (n =3); ns, not significant. (B) Proportion of colonies determined by microscopicanalysis. (C) Representative photographs of BFU-E–type colonies (Upper;without TGF-β1) and CFU-E–type colonies (Lower; with TGF-β1) from CD71+

MEPs. (D) Each FACS-purified fraction was cultured for 4 d in serum-freemedium supplemented with SCF, TPO, and EPO, ± TGF-β1 (2 ng/mL), and thenassayed for GPA expression. Black, unstained control; blue, without TGF-β1;red, with TGF-β1. Similar results were reproduced in two independent ex-periments. (E ) Representative photographs of GPA+ cells obtained fromCD71− CD105− MEPs (Left) and CD71int/+ CD105+ EPs (Right) in the absence(Upper) and presence (Lower) of TGF-β.

HSC

EP

E-MEP

MkP

Pre-MEP

GMP

CD71-CD105- CLP

MPP

T, B, NK-cells

GranulocytesMonocytes

Erythrocytes

Platelets

Flt3

CD71Flt3

OriginalMEP

OriginalCMP

CD105

IL-3Rα

Fig. 6. Proposed model of the developmental pathway in human myeloidprogenitors. The originally defined human CMP contained the unbiased (orslightly GM-biased) CD71− CD105− fraction and MegE-biased CD71+ CD105−

fraction (pre-MEPs; yellow). The hMEP downstream of these fractions con-tained a CD71+ CD105− fraction of E-MEPs (yellow) and a CD71int/+ CD105+

fraction of EPs (orange), respectively. NK, natural killer.

Mori et al. PNAS | August 4, 2015 | vol. 112 | no. 31 | 9641

CELL

BIOLO

GY

Dow

nloa

ded

by g

uest

on

Janu

ary

26, 2

022

CD105− E-MEPs gave rise to EPs but fewer CD71− CD105−

MEPs, whereas CD71int/+ CD105+ EPs never generated CD105−

MEPs. These findings indicate that the main stream of E-lineagedevelopment continues from CD71− CD105− CMPs, via CD71+

CD105− CMPs and CD71+ CD105− MEPs, to EPs (Fig. 6). MegKprogenitors may be isolatable somewhere in the latter pathway.There are some unclarified issues at this point. (i) Is the

CD71int/+ CD105+ hEP the earliest stage of unipotent E-lineagecells? The expression of CD36 precedes that of CD105 (Fig. 1); thus,it could be that CD71+ CD36+ CD105− cells within the originalCMP and/or MEP fractions are earlier committed E progenitors(e.g., pre–BFU-E), although CD36 is reported to be expressed alsoon MegKs, monocytes (49), and/or a part of CD13+ CD133+ bipo-tent myeloerythroid progenitors (50). Moreover, as shown in Fig. 1B,the expression level of CD36 in CMP subpopulations is quite low,and thus additional cell-surface markers may be required to furtherinvestigate this issue. (ii) Is the E-lineage developmental pathwaycommon between fetal and adult hematopoiesis? We found that farfewer EPs existed in cord blood (CB) than in BM (Fig. S1). Thisfinding may suggest the immaturity of CB erythroid cells, and couldpartially explain the delayed recovery of red blood cells and plateletsresulting in the increased frequency of blood transfusions frequentlyseen among patients who received CB transplants compared withallogeneic mobilized peripheral blood stem cell transplants. Thatbeing said, both HSC numbers and HSC engraftment potential areother potential contributing variables (51). CB might possess un-known E-lineage developmental pathways different from adult he-matopoiesis. (iii) Are hEPs involved in physiological erythropoiesisin vivo? We have not tested the in vivo lineage potential of isolatedhEPs in a xenotransplantation model (52) because of their smallnumbers and limited proliferation capacity. These subsets of hMEPand hEP populations in patients with unexplained anemia, MDS, orPV will need to be analyzed in comparison with healthy individuals.(iv) How can one isolate MegK lineage-committed progenitors?We found that the expression level of CD41, a well-establishedmarker for MegK progenitors both in mouse (9) and human(13, 25), was up-regulated with MegE differentiation and down-regulated in CD105+ EPs (Fig. S2 B and C). To isolate MegK-committed progenitors at the CMP or MEP level, a marker otherthan or in addition to CD41 is necessary. Recently, MegK-biasedHSCs (53) and MegK-committed progenitors with long-term repo-pulating capacity (54) within the mouse HSC compartment havebeen reported. In humans as well, a portion of MegK progenitorsmight develop directly from earlier stem cells/progenitors such asHSCs and MPPs, bypassing the CMP or MEP stage.Although the MegK developmental pathway remains unde-

termined, the use of key TFs, at least in the E lineage, appearsto be well-preserved between human and mouse (Fig. S5). TheCD71− CD105− CMPs express low levels of both MegE-related(GATAs, KLF-1, Fli-1) and GM-related (C/EBPα, PU.1) TFs,with the up-regulation of the former and down-regulation of thelatter as the CMP differentiates downstream toward the MEP viaan intermediate CD71+ CD105− CMP stage. Subsequent KLF-1up-regulation accompanied by Fli-1 down-regulation may becritical for the E-lineage fate determination at the branch-pointof the MegK versus E lineage.In summary, we have identified hEPs as a CD71int/+ CD105+

fraction of previously defined hMEPs. This population could be auseful tool for studying physiological and pathological erythroid

cell development, and should be analyzed in patients sufferingfrom anemia or erythrocytosis such as in MDS or PV.

Materials and MethodsAntibodies, Cell Staining, and Sorting. The sorting procedures for HSCs andmyeloid progenitor populations that we previously reported (20) wereslightly modified. In brief, bone marrow mononuclear cells, purchased fromAllCells, were first stained with phycoerythrin (PE)-Cy5–conjugated lineageantibodies, including anti-CD3, -CD4, -CD8, -CD10, -CD19, -CD20, -CD11b, -CD14,and -CD56. Cells were then stained with allophycocyanin (APC)-Cy7–conju-gated anti-CD34 (BioLegend), APC-conjugated anti-CD38 (BD Pharmingen),PE-conjugated anti–IL-3Rα (BD Pharmingen), and Pacific blue (PaB)- or FITC-conjugated anti-CD45RA (BioLegend or eBioscience) antibodies. CMPs, GMPs,and MEPs were isolated as Lin− CD34+ CD38+ IL-3Rα+ CD45RA−, Lin− CD34+

CD38+ IL-3Rα+ CD45RA+, and Lin− CD34+ CD38+ IL-3Rα−/lo CD45RA− pop-ulations, respectively. To sort EPs and MegK progenitors, FITC-conjugatedanti-CD71 (BioLegend) and biotinylated anti-CD105 (endoglin; BioLegend)antibodies followed by streptavidin-PE-Cy7 (BD) were added. Pre-EMPs, E-MEPs,and EPs were purified as Lin− CD34+ CD38+ IL-3Rα+ CD45RA− CD71+ CD105−,Lin− CD34+ CD38+ IL-3Rα−/lo CD45RA− CD71+ CD105−, and Lin− CD34+ CD38+

IL-3Rα− CD45RA− CD71+ CD105+ populations, respectively. MegK progenitorswere enriched within the Lin− CD34+ CD38+ IL-3Rα−/lo CD45RA− CD71− CD105−

fraction. To evaluate Flt3 expression, costaining of CD105 was omitted due totechnical difficulties. Dead cells were excluded by propidium iodide staining.All sorting and analyses were performed on a three laser-equipped FACSAria II(BD Biosciences). The automatic cell-deposition system was used for single-cellassays. FACS data were analyzed with FlowJo software (Tree Star).

Cell Culture. For short-term liquid cultures, purified populationswere suspendedon 12-well plates with the following medium: Iscove’s modified Dulbecco’smedium (Life Technologies) supplemented with 20% (vol/vol) FCS, antibiotics,20 ng/mL human recombinant SCF, 20 ng/mL GM-CSF, 4 U/mL EPO, and 20 ng/mLTPO (R&D Systems). IL-3 (20 ng/mL) was added when CMPs were cultured.For clonogenic analysis of myeloid progenitors, cells were cultured for 14 d inIscove’s Modified Dulbecco’s Medium (IMDM)-based methylcellulose medium(MethoCult GF H4434; StemCell Technologies), which contained FCS, BSA,2-mercaptoethanol, recombinant SCF, IL-3, GM-CSF, and EPO, with an addi-tional 20 ng/mL TPO. For the MegK assay, serum-free medium (StemSpanTMSFEM II; StemCell Technologies) was used.

All cultures were incubated in a humidified chamber in 5% CO2. Colonies werescored and picked up formaking cytospin preparations to define cell components.

Gene Expression Analysis. Total RNA was extracted from purified progenitorpopulations using TRIzol reagent (Life Technologies) according to themanufacturer’s protocol. All RNA samples were reverse-transcribed withOligo(dT) primers using the SuperScript III First-Strand Synthesis System(Invitrogen/Life Technologies). Quantitative real-time PCR assays were per-formed with the 7900HT Fast Real-Time PCR System (Life Technologies). Allspecific primers and probes were purchased from inventoried stocks ofTaqMan Gene Expression Assays (Applied Biosystems/Life Technologies).GAPDH transcripts were simultaneously amplified as an internal standard forquantification. All samples were analyzed in triplicate.

Statistical Analysis. The unpaired two-tailed Student t test was applied to allpairwise comparisons of mean values after F-test evaluation of variance. Allstatistical analyses were performed with Prism 5 software (GraphPad).

ACKNOWLEDGMENTS. We thank T. Storm and L. Jerabek for laboratorymanagement and T. J. Naik for technical assistance. This study was sup-ported by fellowships from the Japan Society for the Promotion of Science(to Y.M.) and grants from the National Cancer Institute and National Heart,Lung, and Blood Institute of the National Institutes of Health (R01CA086065 and U01 HL099999; to I.L.W.), California Institute for RegenerativeMedicine (RT2-02060; to I.L.W.), and Leukemia & Lymphoma Society (700709;to I.L.W.).

1. Orkin SH, Zon LI (2008) Hematopoiesis: An evolving paradigm for stem cell biology.Cell 132(4):631–644.

2. Spangrude GJ, Heimfeld S, Weissman IL (1988) Purification and characterization ofmouse hematopoietic stem cells. Science 241(4861):58–62.

3. Morrison SJ, Weissman IL (1994) The long-term repopulating subset of hemato-poietic stem cells is deterministic and isolatable by phenotype. Immunity 1(8):661–673.

4. Seita J, Weissman IL (2010) Hematopoietic stem cell: Self-renewal versus differentia-tion. Wiley Interdiscip Rev Syst Biol Med 2(6):640–653.

5. Pang WW, et al. (2013) Hematopoietic stem cell and progenitor cell mechanisms inmyelodysplastic syndromes. Proc Natl Acad Sci USA 110(8):3011–3016.

6. Vardiman J, Hyjek E (2011) World Health Organization classification, evaluation, andgenetics of the myeloproliferative neoplasm variants. Hematology Am Soc HematolEduc Program 2011:250–256.

7. Akashi K, Traver D, Miyamoto T, Weissman IL (2000) A clonogenic common myeloidprogenitor that gives rise to all myeloid lineages. Nature 404(6774):193–197.

8. Manz MG, Miyamoto T, Akashi K, Weissman IL (2002) Prospective isolation of humanclonogenic commonmyeloid progenitors. Proc Natl Acad Sci USA 99(18):11872–11877.

9642 | www.pnas.org/cgi/doi/10.1073/pnas.1512076112 Mori et al.

Dow

nloa

ded

by g

uest

on

Janu

ary

26, 2

022

9. Pronk CJ, et al. (2007) Elucidation of the phenotypic, functional, and molecular to-pography of a myeloerythroid progenitor cell hierarchy. Cell Stem Cell 1(4):428–442.

10. Baum CM, Weissman IL, Tsukamoto AS, Buckle AM, Peault B (1992) Isolation of acandidate human hematopoietic stem-cell population. Proc Natl Acad Sci USA 89(7):2804–2808.

11. Rogers CE, Bradley MS, Palsson BO, Koller MR (1996) Flow cytometric analysis ofhuman bone marrow perfusion cultures: Erythroid development and relationshipwith burst-forming units-erythroid. Exp Hematol 24(5):597–604.

12. Terstappen LW, Huang S, Safford M, Lansdorp PM, Loken MR (1991) Sequentialgenerations of hematopoietic colonies derived from single nonlineage-committedCD34+CD38− progenitor cells. Blood 77(6):1218–1227.

13. Novershtern N, et al. (2011) Densely interconnected transcriptional circuits control cellstates in human hematopoiesis. Cell 144(2):296–309.

14. Pierelli L, et al. (2001) CD105 (endoglin) expression on hematopoietic stem/progenitorcells. Leuk Lymphoma 42(6):1195–1206.

15. Bühring HJ, et al. (1991) Endoglin is expressed on a subpopulation of immatureerythroid cells of normal human bone marrow. Leukemia 5(10):841–847.

16. Lammers R, et al. (2002) Monoclonal antibody 9C4 recognizes epithelial cellular ad-hesion molecule, a cell surface antigen expressed in early steps of erythropoiesis. ExpHematol 30(6):537–545.

17. Rokhlin OW, Cohen MB, Kubagawa H, Letarte M, Cooper MD (1995) Differentialexpression of endoglin on fetal and adult hematopoietic cells in human bone mar-row. J Immunol 154(9):4456–4465.

18. Mirabelli P, et al. (2008) Extended flow cytometry characterization of normal bonemarrow progenitor cells by simultaneous detection of aldehyde dehydrogenase andearly hematopoietic antigens: Implication for erythroid differentiation studies. BMCPhysiol 8:13.

19. Machherndl-Spandl S, et al. (2013) Molecular pathways of early CD105-positive ery-throid cells as compared with CD34-positive common precursor cells by flow cyto-metric cell-sorting and gene expression profiling. Blood Cancer J 3:e100.

20. Majeti R, Park CY, Weissman IL (2007) Identification of a hierarchy of multipotenthematopoietic progenitors in human cord blood. Cell Stem Cell 1(6):635–645.

21. Okumura N, Tsuji K, Nakahata T (1992) Changes in cell surface antigen expressionsduring proliferation and differentiation of human erythroid progenitors. Blood 80(3):642–650.

22. de Wolf JT, Muller EW, Hendriks DH, Halie RM, Vellenga E (1994) Mast cell growthfactor modulates CD36 antigen expression on erythroid progenitors from humanbone marrow and peripheral blood associated with ongoing differentiation. Blood84(1):59–64.

23. Kikushige Y, et al. (2008) Human Flt3 is expressed at the hematopoietic stem cell andthe granulocyte/macrophage progenitor stages to maintain cell survival. J Immunol180(11):7358–7367.

24. Edvardsson L, Dykes J, Olofsson T (2006) Isolation and characterization of humanmyeloid progenitor populations—TpoR as discriminator between common myeloidand megakaryocyte/erythroid progenitors. Exp Hematol 34(5):599–609.

25. Debili N, et al. (1992) Expression of CD34 and platelet glycoproteins during humanmegakaryocytic differentiation. Blood 80(12):3022–3035.

26. Clay D, et al. (2001) CD9 and megakaryocyte differentiation. Blood 97(7):1982–1989.27. Nakorn TN, Miyamoto T, Weissman IL (2003) Characterization of mouse clonogenic

megakaryocyte progenitors. Proc Natl Acad Sci USA 100(1):205–210.28. Ng AP, et al. (2012) Characterization of thrombopoietin (TPO)-responsive progenitor

cells in adult mouse bone marrow with in vivo megakaryocyte and erythroid po-tential. Proc Natl Acad Sci USA 109(7):2364–2369.

29. Cheifetz S, et al. (1992) Endoglin is a component of the transforming growth factor-beta receptor system in human endothelial cells. J Biol Chem 267(27):19027–19030.

30. Krystal G, et al. (1994) Transforming growth factor beta 1 is an inducer of erythroiddifferentiation. J Exp Med 180(3):851–860.

31. Zermati Y, et al. (2000) Transforming growth factor inhibits erythropoiesis byblocking proliferation and accelerating differentiation of erythroid progenitors. ExpHematol 28(8):885–894.

32. Hino M, Tojo A, Miyazono K, Urabe A, Takaku F (1988) Effects of type beta trans-forming growth factors on haematopoietic progenitor cells. Br J Haematol 70(2):143–147.

33. Akel S, Petrow-Sadowski C, Laughlin MJ, Ruscetti FW (2003) Neutralization of auto-crine transforming growth factor-beta in human cord blood CD34(+)CD38(−)Lin(−)cells promotes stem-cell-factor-mediated erythropoietin-independent early erythroidprogenitor development and reduces terminal differentiation. Stem Cells 21(5):557–567.

34. Böhmer RM (2004) IL-3-dependent early erythropoiesis is stimulated by autocrinetransforming growth factor beta. Stem Cells 22(2):216–224.

35. Iwasaki H, et al. (2006) The order of expression of transcription factors directs hier-archical specification of hematopoietic lineages. Genes Dev 20(21):3010–3021.

36. Mori Y, et al. (2009) Identification of the human eosinophil lineage-committed pro-genitor: Revision of phenotypic definition of the human common myeloid pro-genitor. J Exp Med 206(1):183–193.

37. Fujiwara Y, Browne CP, Cunniff K, Goff SC, Orkin SH (1996) Arrested development ofembryonic red cell precursors in mouse embryos lacking transcription factor GATA-1.Proc Natl Acad Sci USA 93(22):12355–12358.

38. Mancini E, et al. (2012) FOG-1 and GATA-1 act sequentially to specify definitivemegakaryocytic and erythroid progenitors. EMBO J 31(2):351–365.

39. Tsang AP, Fujiwara Y, Hom DB, Orkin SH (1998) Failure of megakaryopoiesis andarrested erythropoiesis in mice lacking the GATA-1 transcriptional cofactor FOG.Genes Dev 12(8):1176–1188.

40. Frontelo P, et al. (2007) Novel role for EKLF in megakaryocyte lineage commitment.Blood 110(12):3871–3880.

41. Bouilloux F, et al. (2008) EKLF restricts megakaryocytic differentiation at the benefitof erythrocytic differentiation. Blood 112(3):576–584.

42. Hart A, et al. (2000) Fli-1 is required for murine vascular and megakaryocytic devel-opment and is hemizygously deleted in patients with thrombocytopenia. Immunity13(2):167–177.

43. Starck J, et al. (2003) Functional cross-antagonism between transcription factors FLI-1and EKLF. Mol Cell Biol 23(4):1390–1402.

44. Wu H, Liu X, Jaenisch R, Lodish HF (1995) Generation of committed erythroid BFU-Eand CFU-E progenitors does not require erythropoietin or the erythropoietin re-ceptor. Cell 83(1):59–67.

45. Lin CS, Lim SK, D’Agati V, Costantini F (1996) Differential effects of an erythropoietinreceptor gene disruption on primitive and definitive erythropoiesis. Genes Dev 10(2):154–164.

46. Wendling F (1999) Thrombopoietin: Its role from early hematopoiesis to plateletproduction. Haematologica 84(2):158–166.

47. Sivertsen EA, et al. (2006) PI3K/Akt-dependent Epo-induced signalling and targetgenes in human early erythroid progenitor cells. Br J Haematol 135(1):117–128.

48. Berthier R, Valiron O, Schweitzer A, Marguerie G (1993) Serum-free medium allowsthe optimal growth of human megakaryocyte progenitors compared with humanplasma supplemented cultures: Role of TGF beta. Stem Cells 11(2):120–129.

49. Edelman P, et al. (1986) A monoclonal antibody against an erythrocyte ontogenicantigen identifies fetal and adult erythroid progenitors. Blood 67(1):56–63.

50. Chen L, Gao Z, Zhu J, Rodgers GP (2007) Identification of CD13+CD36+ cells as acommon progenitor for erythroid and myeloid lineages in human bone marrow. ExpHematol 35(7):1047–1055.

51. Solh M, Brunstein C, Morgan S, Weisdorf D (2011) Platelet and red blood cell utili-zation and transfusion independence in umbilical cord blood and allogeneic pe-ripheral blood hematopoietic cell transplants. Biol Blood Marrow Transplant 17(5):710–716.

52. Ishikawa F, et al. (2005) Development of functional human blood and immune sys-tems in NOD/SCID/IL2 receptor gamma chain(null) mice. Blood 106(5):1565–1573.

53. Sanjuan-Pla A, et al. (2013) Platelet-biased stem cells reside at the apex of the hae-matopoietic stem-cell hierarchy. Nature 502(7470):232–236.

54. Yamamoto R, et al. (2013) Clonal analysis unveils self-renewing lineage-restrictedprogenitors generated directly from hematopoietic stem cells. Cell 154(5):1112–1126.

Mori et al. PNAS | August 4, 2015 | vol. 112 | no. 31 | 9643

CELL

BIOLO

GY

Dow

nloa

ded

by g

uest

on

Janu

ary

26, 2

022