Bravo. J Burckhardt. T Curran R Muller, EMBO J. 4 , 1 193 (1 985)

10 X Y . Fu and J.-J Zhang, Cell74. 1135 (1993) 11. S. Ruff-Jatnson, K. Chen. S Cothen, Science 261,

1733 (1993) S Ruff-Jamison et a/. , J. Blol. Ciiem. 269 21933 (1994); H B Sadowski and M. Z GII- man. Nature 362, 79 (1993).

12 Z Zhong. Z. Wen, J E Darnel Jr . Science 264. 95 (1 994).

13 A431 cells from the AmericanType Culture Collection (ATCC CRL-1555) were grown in Dulbecco's modl- fled Eagle's medium (DMEM) witt? calf serum (1 0%) HT29 cells (ATCC HTB-38) were grown In McCoy's 5Amedum with fetal bovine serum (FBS) (1 0%). WDr cells (ATCC CLL-218) were grown in m n m a essen- tial m e d ~ u ~ n w~ th FBS (10%). Human recombinant EGF was obtaned from Gbco BRL. FN--1 from Gen- entecth and'ant~bodes to STAT1 and STAT3 from Santa Cruz Biotecthnology Whole-cell extracts were prepared as descrbed (10) by ysis of cells In 20 mM Hepes (pH 7.9) buffer containing 0.2% NP-40. 10% glycerol. 400 mM NaCl 0 1 mM EDTA. 1 mM ditho- threitol. 1 mM sod~um vanadate. 0 5 mM pthenyl- metthylsulfonyl fluoride, and aprotinin, leupeptin. and pepstat~n (1 p.g/m of eacth). For EMSAs, double- stranded oligodeoxynucleot~de probes were end-a- beled w~ th [y-3iP]ATP and portions equ~valent to 20.000 cpm were used per reaction. B~nding reac- tons were performed in a total volume of 15 p.1 in 10 mM Hepes (pH 7.9). 0 1 mM EDTA. 5% glycerol. poly(deoxy~nos~ne-deoxycytd~ne) (50 p.g/ml, Pharma- cia), and 0.01 % NP-40 Extracts were Incubated for 10 m n on ice. antbody (1 /*I) was added followed by an add~tional 30-m~n.incubation on ice, and tthen the end-labeled DNA probe was added and Incubated for 20 m n at room temperature. Complexes were sepa- rated on nondenaturing acrylam~de gels (6%) In 0 5X tr~s-borate EDTA and detected by autoradiography.

14. Y E. Chln and X.-Y Fu, unpublisthed data. 15 B J. Wagner, T E Hayes, C J Hoban. B H. Coch-

ran. EbABO J. 9, 4477 (19901 16 W. S El-De~ry et a1 . Cancer Res. 54, 11 69 (1 994) 17. H B Sadowski. K Sthua~. J E. Darnel J r . M Z

Glman, Sclence 261. 1739 (1 993). 18. Total cellular RNA was prepared by the guanidinium

thiocyanate-CsCl procedure. RNA (5 mg) was sep- arated on a 1 0% agarose-formaldehyde gel and transferred onto a nylon membrane (Zeta-Probe. Bio-Rad) The flter was stained witt? ~netthylene blue ID L Herrin and G W Schmidt. B~otechnioues 6. i 9 6 (1988)l and then thybridized at 65°C In 0 25 M Na,PO, (pH 7 2), 7% SDS, and 1 mM EDTA. The wast? was performed at 65°C In 0.04 M Na,PO, (pH 7 2) and 1% SDS. The probe was preparedby abel- n g a Stu I-Xtho I fragment (1.9 kb) of human p21 complementary DNA (3) w~ th the use of a random- prmed DNA labeling kit (Boethringer Mannhem)

19. M. Kitagawa. W.-C S. Su, X.-Y Fu, unpubshed data

20. Z -H. You and X Y . Fu. unpublished data 21. R. McKendry et a1 . Proc. Natl Acad SCI U S.A 88.

11455 (1991), M Muller e ta / . , EMBO J. 12 4221 ( 1 993)

22 U3A cells (21) were transfected w ~ t h 0 3 p.g of pST- neoB inearzed with Xho I [K. Katot?, Y Takathasthi, S. hay ash^, H. Kondoh. CellStruct Funct. 12, 575 (1 987)] and 15 l ~ g of either pSG5 (Stratagene) or pSG91 (10) linearized with S a I by the calc~um phosphate metthod [F. M Ausubel et a / . . Culrent Protocols in Molecular Biology f Wileylntersc~ence. New York, 1987)] G418-resistant cells were se- lected and ma~ntaned n medium contanng G418 (700 p.g/ml).

23 The cells (2 x 10" cells In a six-well microculture plate) treated w~ th EGF or IFN-y were incubated with ["Hlthymidine (5 p.Ci.ml) for 6 hours. Cells were then washed twce w~ th phosphate-buffered same har- vested onto the glass fliers and added to vials con- taning scint~llat~on fluid for liqu~d-sc~ntillation mea- surement of 'H incorporation.

24 A two-ayer soft-agar system was prepared by plat- ~ n g 0 5% agar ~n DMEM containing calf serum (20%) in 60-mm dishes [A, l/\i. Hamburger and S E. Sam- on. Science 197, 461 (197711. Cells were passed through a 25-gauge needle and then suspended In a 0.35% agar solut~on In DMEM enriched w~ th calf

serum (20%). Tthe f~nal concentratlon of cells in 0.35% agar solution was 2 5 x 10" cel1s.m In 2 ml(5 X 10" cells) Plates were examined mlned~ately after plating to ensure that a single-cell suspension In agar thad been actheved, and were then Incubated All specmens were plated in trpicate. DMEM (0.2 ml) contanng calf serum (1 0%) was added to each well FN-y was added to the fnal plating mixture just before plating (f~nal concentratlon, 100 ng/ml), but not to control plates FN--1 was then added d a y after plating to maintain the effect Colonies were counted (a colony was defined as a new round ag- gregate of 50 or more cells) on day 8 witt? a phase- contrast inverted mcroscope

25. T. Kwok, T Mok, C H. Menton-Brennan, Cancer Res. 54, 2834 (1 994)

26 Z Fan et a1 . J. Cell Biol 131, 235 (1 995) 27 D. Resn~tzky, N Tiefenbrun, H Ber~ss A. K~mchi,

Proc Natl Acad. Sci. U.S.A 89, 402 (1992). 28. We thank G. R Stark and I. M. Kerr for the U3A cell

line, Y. Xiong for p21 cDNA: Genentech for IFN--1; W.-F Tao for some of tthe cells. D. Stern, D Breg- man, and H. Zthang for crtcal readng of the manu- script and helpful discussions: N Bennett for assist- ance in preparing tthe manuscript; and colleagues In our department for support. Supported by grants from tthe American Cancer Soc~ety (NP883), NIH (R01 Al 34522) and the Counc~l for Tobacco Re- search (4238) to X Y . F Y E C. is supported by a N H tra~ning grant.

26 January 1996, accepted 6 Marct? 1996

In Vitro Development of Primitive and Definitive Erythrocytes from Different Precursors

Toru Nakano,* Hiroaki Kodama, Tasuku Honjo

During mouse embryogenesis the production of "primitive" erythrocytes (EryP) precedes the production of "definitive" erythrocytes (EryD) in parallel with the transition of the hematopoietic site from the yolk sac to the fetal liver. On a macrophage colony-stimulating factor-deficient stromal cell line OP9, mouse embryonic stem cells were shown to give rise to EryP and EryD sequentially with a time course similar to that seen in murine ontogeny. Studies of the different growth factor requirements and limiting dilution analysis of precursor frequencies indicate that most EryP and EryD probably developed from different precursors by way of distinct differentiation pathways.

Erythropoiesls originates in the yolk sac, then migrates tcl the fetal liver during mcluse embryogenesis. EryP and EryD, wllich are produced in the yolk sac and the fetal liver, resvectivelv, have d~st inct n o r - , , ,

pholog~cal and biochemical characteristics ( 1 , 2). Whether these two cell types devel- op frclm a single cclmmcln hematclpoiet~c precursor or not has been the subject of controversy (3 ,4) . T o address this question, we used the In vitro differentiation induc- tion system of emhryclnic stem (ES) cells to helnatopoiet~c cells (5).

Two waves of erythro~d cell production were observed when D3 ES cells were cocul- tured with OP9 stromal cells (Fig. 1A) (5- 8). T h e first n7a\7e of erythrclpoiesis appeared at day 6 of the induct~on, and all of the dav 7 erythroid lineage cells were large-nucleated cells rnorphologi~al l~ identical tcl EryP (Fig. 2A). T h e number of erythrclld lineage cells s~~ddellly decreased at days 8 and 9 to less than one-flit11 of that at day 7. Subsequently, the second wave of ervthroid lineaee cells appeared around day 16 w ~ t h a peaka t day 14; these cells were small-nucleated ervthro- blasts or enucleated mature blood cells mor-

T Nakano and T. Honjo Department of Medca Chem- istry, Faculty of Med~c~ne. Kyoto Un~versity. Yoshida Sakyo-ku, Kyoto 606, Japan. H Kodama Department of Anatomy, Ohu Un~vers~ty School of Dentistry, tom tam act?^, Koriyama, Fukushima 963, Japan.

"Present address: Department of Molecular Cell Biology, Researct? Institute for Microb~al Dseases. Osaka Unver- sity, Yamada-Oka, Suta, Osaka 565, Japan

pholog~cally identical to EryD (Fig. 2B). In agreement with the report that EryP contam not onlv ernbrvon~c 1- and e-e lohn but also adult a-globin, whereas EryD contain only adult a- and P-globins (9) , day 7 erythroid lineage cells were positive for staining with ant~bodies aeamst embr\jcln~c as well as - against ailult hemoglob~ns, rvhereas day 14 ervthro~d cells were nositive for staining with a1;tihodies against 'adult llemclglok,il< only (Fig. 2, C to F) (1 0-1 2). Expression of j - , a-, and e-globin IIIRNA in day 7 erythrocytes

Days after differentiation induction

Fig. 1. Development of erythroid Ineage cells dur- Ing differentiaton nducton of D3 ES cells on OP9 stromal cells, and effects of antl-c-KI~ (Ack2) and erythropoietin (EPO) on the development (24). Dif- ferentiation induction of ES cells was done wthout (A) or with (6) the additon of exogenous human recombinant EPO (1 0 U/m) (8). Data are shown In the absence (0) or presence (0) of Ack2 (1 0 pg/ mli.

SCIENCE VOL 272 3 M.4Y I996

and of a- and P-major-globin mRNA in day 14 erythrocytes was confirmed by reverse transcriptase-polymerase chain reaction (RT-PCR) (Fig. 2G) (13). These results show that erythroid lineage cells produced in the first and the second waves of erythropoi- esis are EryP and EryD, respectively. During the differentiation induction of ES cells, EryP and EryD first became detectable at day 6 and 10, respectively. The temporal pattern of the appearance of EryP and EryD on the stromal cells is similar to that of erythropoi- esis during mouse ontogeny, in which EryP and EryD appear at days 8 and 12, respec- tively (1 ).

To test whether EryP and EryD require different growth factors for their develop- ment and survival, we examined the in- volvement of c-Kit ligand and erythropoie- tin (EP0)-induced signaling (Fig. 1). Ad- dition of antibody to c-Kit (anti-c-Kit) (Ack2) (14) did not affect EryP develop- ment but completely blocked EryD devel- opment (Fig. 1A). c-Kit signaling seems to be necessary for the development or main- tenance (or both) of multipotential hema- topoietic precursors from ES cells in this system because not only EryD but also other myeloid lineage cells did not develop in the presence of AckZ at day 14 (15). Because EPO mRNA was not detectable in the cocultured cells by RT-PCR (15), both EryP and EryD developed independently of EPO. This conclusion is consistent with the recent report that EryD precursor formation and impaired EryP production were ob- served in EPO-targeted and EPO receptor- targeted mice (16). In the presence of EPO, the type of erythroid lineage cells as as- sessed by immunostaining was exclusively EryP before day 9, and the percentage of

EryP gradually decreased to -70% at day 11, -10% at day 12, and less than 1% after day 13. Even though c-Kit signaling was blocked by AckZ, -80% of day 14 erythroid lineage cells were EryD in the presence of EPO (Fig. 1B). This impaired but notable production of EryD is comparable to the phenotype of the mice bearing null muta- tion of c-Kit (1 7). . ,

Further investigation to distinguish pre- cursors for ErvP and ErvD revealed that END precursors could be easily isolated and sepa- rated from EryP precursors simply by collect- ing cells adherent to stroma cells in the EPO-containing culture at day 6 (Table 1). The vast maioritv of ErvD develo~ed from , , the adherent cell population, whereas EryP were mostly recovered from the nonadherent cell population. Adherent cells grew by day 8 to small round cell clusters consisting of immature hematopoietic cells on OP9 cells (5). When 120 individual, well-separated, small round cell clusters were nicked at dav 7 and transferred separately into methylcellu- lose cultures containing interleukin-3 and EPO without dispersing the cells (5, 18, 19), 11 1 clusters produced blood cell col- onies, including 91 erythroid mixed colo- nies (EryD plus at least one other lineage cell), 7 EryD colonies, 10 nonerythroid colonies, and 3 unidentified cell-type col- onies. Immunostainine showed that none " of the colonies contained embryonic he- moglobin-positive erythrocytes. Thus, most EryD developed from multipotential hematopoietic progenitors. These data suggest that EryP and EryD are produced by distinct differentiation programs.

The time course of the appearance of precursors of EryP and hematopoietic clus- ters, from which EryD were derived as dis-

Fig. 2. Morphology of Day 7 Day 14 erythroid lineage cells and globin gene expres- sion at days 7 and 14 of the differentiation induc- tion in the absence of EPO. Cytospin prepara- tions of the cells harvest- ed from day 7 (A, C, and E) or day 14 (8, D, and F) were stained either with May-Giemsa (A and B) or by the PAP immunoen- zymatic method (C to F). Polyclonal anti-mouse embryonic hemoglobin (C and D) and polyclonal anti-mouse adult hemo- globin (E and F) were used as primary antibod- ies (1 1). Expression of embryonic globin genes (5 and E ) and adult globin aenes (a- and B-maior) has examined 'by RT: PCR (G) (13).

cussed above, was examined by limiting dilution analysis (Table 2). Substantial numbers of precursors of EryP and hemato- poietic clusters appeared at days 3 and 4 of the induction, respectively, and the number of precursors increased with time. We then tested whether common precursors for EryP and EryD existed in cultured ES cells at early stages of differentiation induction, that is, day 3 and day 4. Most of the hema-

Table 1. Development of EryP and EryD from day 6 adherent or nonadherent cells in the culture containing EPO. Cells were examined at day 8 and day 13 in the culture containing EPO. Cocul. tured D3 ES cells were trypsinized and transferred onto fresh OP9 cells at day 5. On the next day, cells nonadherent to stromal cells were collected by washing gently three times with culture medi- um and transferred onto fresh OP9 cells as non- adherent population. Culture medium was added to the remaining adherent cell layers, and cocul- tures were continued as adherent population. A control unseparated cell population received no treatment except that total cells were transferred to fresh OP9 cell layers at day 10. Differentiation induction was carried out in the presence of hu- man recombinant EPO (10 U/ml). Erythroid lin- eage cells were identified by May-Giemsa stain- ing, and the data are shown as the mean + SE. Day 5 cells (lo5) were used for separated and unseparated cell populations.

Erythroid cells (XI 0-4) Cell fractions

at day 6 Day 8 Day 13 EryP/ (EryP) (EryD) EryD

Unseparated 12.2 ? 0.4 271.1 + 13.3 0.045 Nonadherent 7.0 5 0.6 21.5 + 2.8 0.326 Adherent 1.6 + 0.2 225.4 5 16.9 0.007

Table 2. Time course of increase in the number of EryP precursors and hematopoietic cell cluster precursors generated from 1 O4 D3 ES cells (25). Cocultured D3 ES cells were harvested by trypsinization 3,4, and 5 days after the initiation of the differentiation induction in the presence of EPO. Sixty, 150, 300, and 600 induced cells per well were transferred onto OP9 cell layers of four 24-well plates in the presence of EPO. Three days after the transfer, dianisidine solution was added to the media (26), and dianisidine-positive nonad- herent cells were counted as EryP. Thereafter, culture media were carefully removed and adher- ent small round cell clusters were counted.

Number of precursors

&per- Precursors of at iment

Day 3 Day 4 Day 5

1 Primitive 270 2000 8500 erythrocytes

Hematopoietic <I0 1 100 4800 cell clusters

2 Primitive 250 1100 12,500 erythrocytes

Hematopoietic <I 0 1100 4300 cell clusters

SCIENCE VOL. 272 3 MAY 1996

Table 3. Low frequency for generation of both EryP and hematopoietic cluster precursors from day 3 or day 4 cultured ES cells. Experments were done essentially as descrbed in Table 2. Numbers of the wells positive for EryP precursor, for hematopoietc cluster precursors, or for both were counted after either 60 or 150 induced cells per well were seeded n 192 wells (27).

No. of postive wells Days after No, of cells contanng precursors of Estmated

dfferentation seeded per coincidence of

induction well Primitive Hematopoietic two independent precursors erythrocytes cell clusters

Day 3 60 35 150 76

Day 4 60 30 150 54

topoiesis-positive wells contained either waves of erythroid cell diffentiation in EryP or adherent slnall round cell clusters mouse embryos. The transition from embry- (Table 3). The freauencv of the wells con- onic to adult he~noglohin in mouse as well taining both types of precursors was low and as that from embryonic to fetal henloglobi~l close to the expected value calculated on in human 121) mav thus reflect a switchine , , , - the assumption that the tn.o types of pre- of the cell lineage rather than a switching of cursors developed independent17-. Two the globill gene transcription within an ery- types of erythroid precursors for EryP and throid co~nlnitted cell (22). EryD had heen reported, but whether these two types of precursors developed from REFERENCES AND NOTES common committed precursors or not had

1. R. Rugh. The Mouse (Oxford Univ. Press, Oxford, not heen clear (4) . Our data suggest that 1968). the t\vo types of precursors belong to differ- 2. M. A, S. Moore and D. Metcalf, B r J. Haeniatol. 18.

ellt cohorts, although we cannot exclude 279 (I 970): P. M. C. Wong et a/., Blood 62. 1280

the existence of infrequent or transient (1983); M. L. Craig and E. S. Russel, Dev. Biol 10, 191 11 9641. , ,

common precursor cells to both EryP and 3, v, M, Ingram, Natiire 235, 338 (1972). ErvD. 4. P .M . C. Woncl, S-W. Chuna. S M. Reched. D. H. K.

Differel~tiatioll of EryD took longer than Chul, Blood6?. 716 (1986):~. Keller. M. Kennedy,T. Papayannopulou, M. V. Wiles, Mol. Cell. B!ol. 13,

that of EryP, presumably because EryP pre- 473 (I 993). cursors emerqed directlv from committed 5. T. Nakano, H. Kodama, T. Hon~o. Science 265.1098

precursor cells hut EryD emerged hy way of (I 994); T. Nakano, Sem. /mrnc;nol 7, 197 (1 995). 6. T. C. Doetschman. H. E~stetter, M. Katz, W Schmidt, noncommitted multipotential hematopoi-

R, Kemler, J, Einbryol, Exp, Morphol, 87, 27 (1985), etic precursors. The cell fate determination 7, H, Kodama, M. NO^^, S. ~ i ~ , j ~ . S. ~ i ~ h i k ~ ~ ~ , s.~. to EryP or EryD is most likely governed bv Nishkawa, Exp. Hematol. 22, 979 (1994).

, L

illtrillsic differelltiation progralll per se 8. Dffere"taton nducton of ES cells was carred out as descrbed (5). D3 ES cells that iiad been man-

rather than by the microenvironment, he- tained by standard methods (6, 23) n the presence cause this system the same hema- of leukernla inh~b~tory factor and feeder cells were

topoietic microenvironment for both ery- transferred onto confluent OP9 stro~nal cells in six- well plates at a densty of 1 O4 cells per well. Culture

tllroid lineage precursors on cloned OP? ~ n e d a for the mantenance of OP9 cells and for the stromal cells. Recently, Dzierzak and col- differentaton induction were a-MEM (L~fe Technol- leagues (20) reported new insights into he- lnatopoietic cell development during mouse ontogeny: (i) There are t\vo temporal lvaves of the hematopoietic cell prod~ctioll that could produce EryD; (ii) nonself-renetving heinatopoietic progenitors appear first, fol- lowed hy self-renewing lleinatoroietic stem cells; and (iii) these two progenitors de\~el- op fro111 different anatomical sites. Taken together with the present data, these results indicate that erythroid lineage cells del~elop in three [vaves from three distinct tvnes of , L the cells during mouse ontogeny, that is, ( i ) EryP from EryP precursors, (ii) EryD from nonself-renewing multipotential precursors, and (iii) EryD from self-rene\ving stein cells. Considering that the OP? differentiation induction system could not give rise to self- rene\ving he~natopoietic stein cells (15), this system seems to represent the f~rs t tlvo

ogies, Ga~thersburg, MD) supplemented \w~th 20% fetal bovine seruln (Summit, Australia) and standard antibiotcs. The Induced cells \were trypsinized at day 5, and 1 O5 cells per well were transferred onto fresh OP9 cells. Total cells of ndividua \wells \were harvest- ed by vlgorous p~petting and transferred to a fresh OP9 cell layer on Individual wells again at day 10. Erythrod neage cells were counted by 2.7-damin- ofuorene or May-Gemsa stanng or both, and the two counts \were in general agreement (1 1)

9. A. Leder. A. Kuo. M A Shen, P. Leder. Development 116. 1041 (1992)

10 Cytospin spprimens of day 7 or day 14 cells were stained with May-Giemsa solution The specimens \were fixed \with cold inethanol and stained by perox- idase-antiperox~dase (PAP) immunoenzymatic stain- ing kit as recommended by tlie ~nanufacturer (DAKO, Carpinterla, CAI Rabb~t anti-mouse embryonic he- nog glob in (1l300 dilution) (1 1) and rabbit anti-mouse adult hemoglobin (Cappel Research Products, Durham, NC) \were used as the prmary antbodes.

11. Y Mlm~a, T. Atsumi. N. Imai. Y Ikawa. Development 111, 543 (1991)

12 Viltually all of the erythroid lineage cells on days 6 and 7 \were postive for staining with antsera to em- bryonic hemoglobin. In contrast, less than 1% of

erythrod lineage cells contained embryonic hemo- g o b n after day 10 in the absence of EPO. The rela- tive frequency of EryP and EryD at days 8 and 9 could not be determned accurately because of small numbers of total erythro~d lineage cells. M. J. Weiss, G. Keler. S. H. Orkin, Genes Dev. 8, 11 84 (1 994), J. Sambrook, E. F. Fritsch, T. Manats, Molecular Cloniiig: A Laboratoiy Manual (Cold Spring Harbor Laboratory Press, Cold Spr~ng Har- bor, NY, 19891, RNA was prepared \w~th the use of TRlzol (Life Technologies. Gaithersburg. MD). Total RNA (50 ng) was used for frst-strand synthesis by ol~go(dT) prmer. Polymerase chain reacton (PCR) prmers and PCR conditions were as described (13). M Ogawa et a/., J. Exp. Med. 174, 63 (1991). M. Ogawa eta/. , Developnient 11 7, 1089 (1 993). T. Nakano, unpublshed obsewaton. H. Wu, X. Li t , R Jaensch, H. F. Lodish, Cell 83. 59 (1 955) In EPO receptor-targeted mce, a substantal number of EryD did not appear n v~vo, but EryD precursors c o ~ ~ l d produce EryD colonies n vitro by the addition of a growth factor mlxture. K Nocka, et a/ . Genes Dev. 3, 81 6 (1 989). T. Suda, J. Suda, M. Ogawa, J, hie, J. Cell. Physiol. 124, 182 (1 985). Day 4-induced ES cells were trypsinized and trans- ferred onto an OP9 cell layer. Three days after the transfer, ind~v~dual, welseparated, adherent small round cell clusters were pcked and put en bloc into metliylcellulose media contalnng EPO and interleu- kin-3 (18). Five to 7 days later, emerged colonies were pcked. washed, and divded into three por- tions. Cytospin specmens were produced from the three poltions. One pori~on \was stained with May- Grunwad Giemsa, and the two other portons were staned with the PAP method by use of rabbit antl- mouse embryonic i iemogobn or rabbt anti-mouse adult hernoglobin as the primary antibodies (10). A. L. Medvnsky, N. L Samoyina, A. M. Muler. E. A. Dz~erzak. Nature 364, 64 (1993): A. M. Muller, A. Medvnsky, J Stroubouis, F Grosved. E. Dzierzak, Immunity 1 , 291 (1 994); E Dz~erzak and A. Medvn- sky, Trends Genet. 1 1 . 359 (1 995). G. Stamatoyannopoulos, A. W. Nienhuis, P. W. Ma- jerus, H. Varmus, The Molecular Bass of Blood D s - eases (Saunders, Phiadepha, 1994). M. H. L~ndenbaum and F. Grosveld. Genes Dev. 4, 2075 (1990); C. Peschle et a/., Natui-e 313, 235 (1 985). A. L Joyner, Gene Taigeting (IRL, Oxford, 1993). Data are representative of one of three Independent experiments. Numbers of erythro~d cells are the mean of trlplcate plates and are shown as numbers developed from 1 O5 day 5-induced cells. Standard errors are not Indicated but are located w~thln each crcle. In some exoeriments Ack2 11 0 ualnill or hu- , , - , man recombnant' EPO (1 0 U!m) were added. H. Brevk. J. Cell Phys!ol. 78, 73 (1971) M. C. Cooper, J. Levy, L. N. Cantor, P. A. Marks, R. A. R~fkind, Proc. Natl. Acad Sci. U.S.A. 71, 1677 (1 974). Culture plates that \were seeded wit11 day 4-induced cells were examined 4 days after the transfer as descrbed In Table 2 However, appearance of EryP n culture plates seeded w~ th day 3-induced cells was examined inicroscopically 3 days after the transfer wlthout dianis~dine staining. This microscop- ic assay was relatively easy to perform because more than 90% of nonadherent cells were EryP (15). Day 3 cell cultures were cont~nued for another 2 days to examlne subsequent appearance of adherent hema- topoetc cell clusters m~croscop~cally. This step was necessary because a large number of EryP ap- peared that originated from secondarily d~fferentlat- ed nonhematopoietc colonies when \we waited until 5 days after the transfer (15) We thank S. Okazaki for technical ass~stance, S-I Nishikawa for Ack2 antibody to c-Kit, T Atsum for rabbit ant-mouse embryon~c hemoglobn, and Kirin Brewery for human recombinant EPO reagents Supported by grants f ro~n Vehara Memor~al Founda- t~on, Kanae Foundation of Research for Ne\w Med- cine, and the Ministry of Educaton, Science. Sporis and Culture of Japan.

24 August 1995; accepred 16 February 1996

SCIENCE \'OL 272 3 L1AY 1996