Immunity, Vol. 11, 567–578, November, 1999, Copyright 1999 by Cell Press

Identification of Podocalyxin-like Protein 1as a Novel Cell Surface Marker for Hemangioblastsin the Murine Aorta-Gonad-Mesonephros Region

and generation of long-term repopulating hematopoieticstem cells (LTR-HSCs), originates from the intraembry-onic aorta-gonad-mesonephros (AGM) region at 10.5–11.5 dpc (reviewed by Dzierzak et al., 1998). Within 1–3days of their emergence, LTR-HSCs migrate from the

Takahiko Hara,*# Yoh-ko Nakano,*†Minoru Tanaka,* Kazuhiro Tamura,‡Takashi Sekiguchi,*† Ken-ichi Minehata,*Neal G. Copeland,§ Nancy A. Jenkins,§Masaru Okabe,‖ Hiroshi Kogo,‡Yoh-suke Mukouyama,* and Atsushi Miyajima* AGM region to the fetal liver and then emigrate to the

bone marrow just before birth. Lymphopoietic cells and* Institute of Molecular and Cellular Biosciences†Core Research for Evolutional Science and Technology multipotential hematopoietic progenitors are also de-

tected in the paraaortic splanchnopleura (P-Sp) regionThe University of Tokyo1-1-1 Yayoi, Bunkyo-ku of mouse embryos at 7.5–9.5 dpc (Godin et al., 1995;

Delassus and Cumano, 1996), an intraembryonic siteTokyo 113-0032Japan preceding the AGM region. However, LTR-HSCs that

are capable of repopulating lethally irradiated adult mice‡Tokyo University of Pharmacy and Life Science1432-1 Horinouchi, Hachioji-shi have not been found in the P-Sp region. Interestingly,

it was recently reported that LTR-HSCs can be detectedTokyo 192-0355Japan in the yolk sac and the P-Sp region after transplantion

into the livers of busulfan-treated newborn mice (Yoder§Mammalian Genetics LaboratoryABL-Basic Research Program et al., 1997). Therefore, one speculation has been that

LTR-HSCs generated in these sites lack homing capac-NCI-Frederick Cancer Researchand Development Center ity to the bone marrow and that the phenotypic differ-

ences in hematopoiesis between the yolk sac, the P-SpFrederick, Maryland 21702‖Genome Information Research Center region, and the AGM region can be mostly attributed

to the supporting microenvironment. However, it stillOsaka University3-1 Yamadaoka, Suita remains unknown how LTR-HSCs in the yolk sac acquire

full repopulation activity.Osaka 565-0871Japan Early in this century, detailed observations of the early

development of chick embryos led to the hypothesisthat hematopoietic cells and endothelial cells arise froma common precursor termed the hemangioblast (Sabin,Summary1920; Murray, 1932). In the last 5 years, a number ofstudies have provided evidence supporting this hypoth-Recent studies with avian embryos and murine embry-

onic stem cells have suggested that hematopoietic esis. First, a series of elegant grafting experiments usingchicks and quails demonstrated that the splanchnopleu-cells are derived from hemangioblasts, the common

precursors of hematopoietic and endothelial cells. We ral mesoderm is able to generate hematopoietic cellsand endothelium, while the paraxial mesoderm lacks themolecularly cloned podocalyxin-like protein 1 (PCLP1)

as a novel surface marker for endothelial-like cells in hematogenic capacity (Pardanaud et al., 1996). Hemato-genic activity in the former region is regulated by endo-the aorta-gonad-mesonephros (AGM) region of mouse

embryos, where long-term repopulating hematopoi- derm-derived cytokines such as vascular endothelialgrowth factor (VEGF), basic fibroblast growth factoretic stem cells (LTR-HSCs) are known to arise. PCLP11

CD452 cells in the AGM region incorporated acetylated (bFGF), and transforming growth factor b1 (TGFb1),whereas ectodermal factors such as epidermal growthlow-density lipoprotein and produced both hemato-

poietic and endothelial cells when cocultured with OP9 factor (EGF) suppress it in the latter region (Pardanaudand Dieterlen-Lievre, 1999). In the splanchnopleuralstromal cells. Moreover, multiple lineages of hemato-

poietic cells were generated in vivo when PCLP11CD452 mesoderm, cells expressing VEGF receptor 2 (VEGF-R2)cells were injected into neonatal liver of busulfan- were shown to form both hematopoietic and endothelialtreated mice. Thus, PCLP1 can be used to separate colonies (Eichmann et al., 1997). Furthermore, endo-hemangioblasts that give rise to LTR-HSCs. thelial cells in the dorsal aorta generated CD451 hema-

topoietic cells in vivo as evidenced by cell labelingexperiments using DiI-labeled acetylated low-density li-Introductionpoprotein (DiI-Ac-LDL) (Jaffredo et al., 1998).

While similar in vivo grafting experiments are not pos-Development of hematopoiesis proceeds through twosible in mammalian systems, it was found that hemato-distinct steps, i.e., primitive and definitive hematopoie-poietic cells were clustered at the ventral wall of thesis. In mice, primitive hematopoiesis begins in the extra-dorsal aorta in a 5-week-old human embryo (Tavian etembryonic yolk sac at 7.5 days post coitum (dpc) inal., 1996). More recently, Tavian et al. (1999) showedgestation, while definitive hematopoiesis, which is dis-that CD341 cells in the dorsal aorta and vitelline arterytinguished by enucleated erythrocytes, lymphopoiesis,of human embryos are capable of generating hemato-poietic cells in vitro. In mice, Nishikawa et al. (1998b)# To whom correspondence should be addressed (e-mail: thara@

rinshoken.or.jp). recently showed that the cells expressing Flk1 (mouse

Immunity568

counterpart of VEGF-R2) and vascular endothelial cad- AGM culture may contain hemangioblasts that give riseto hematopoietic progenitors in vitro (Mukouyama et al.,herin (VECadherin) in the yolk sac and the P-Sp region

of mouse embryos at 9.5 dpc gave rise to lymphohema- 1998). To test the possibility that the adherent endothe-lial-like cells produced hematopoietic cells, we exam-topoietic cells in vitro. Similarly, Flk11 hematogenic en-

dothelial cells were generated from ES cells in vitro (Choi ined for uptake of DiI-Ac-LDL, which is known to beincorporated only into endothelial cells and macro-et al., 1998; Nishikawa et al., 1998a). The idea that puta-

tive hemangioblasts express Flk1 arose originally from phages (Voyta et al., 1984). As shown in Figure 1, wefirst labeled the endothelial-like cells generated in thethe finding that knockout mice lacking Flk1 exhibited

severe defects in both hematopoiesis and vasculogen- day 6 AGM culture with DiI-Ac-LDL, and the DiI1CD452

cell population was isolated by cell sorting. In coculturesesis in the yolk sac (Shalaby et al., 1995). Also, Flk1-null cells did not contribute to definitive hematopoiesis of these cells with the unlabeled AGM cells, DiI1CD451

hematopoietic cells appeared after 4 days of incubation(Shalaby et al., 1997), although a recent report sug-gested that a significant number of hematopoietic cells (Figure 1). These hematopoietic cells formed colonies

in the CFU-C assay (data not shown), indicating thatcan be induced from Flk1-null ES cells in vitro (Schuhet al., 1999). some hematopoietic progenitor cells were derived from

the DiI1CD452 endothelial-like cells in the AGM primaryIn conformity with the critical role of TGFb1 in theinduction of hematopoiesis and vasculogenesis in chick culture.embryos, knockout mice deficient in TGFb1 also exhib-ited severe defects in both systems (Dickson et al., Molecular Cloning of Mouse PCLP1 as a Possible1995). Furthermore, gene disruption of the SCL/Tal-1 Hemangioblast Antigentranscription factor caused defects in both hematopoie- To define hemangioblasts more precisely, we attemptedsis and vasculogenesis that were similar to those of Flk1 to obtain a specific antibody directed against heman-knockout mice (Porcher et al., 1996; Visvader et al., gioblasts. By repeating the passage of adherent cells of1998). Mutant zebrafish that are devoid of the SCL/Tal-1 the AGM culture in the presence of OSM, we establishedgene also showed similar defects (Liao et al., 1998), and a novel OSM-dependent endothelial-like cell line, LOforced expression of SCL/Tal-1 in zebrafish embryos (Y. M. et al., unpublished data). LO cells exhibit charac-resulted in the overproduction of hematopoietic and teristics very similar to those of endothelial-like cells invascular cells (Gering et al., 1998). Taken together, these the AGM culture, such as endothelial-like morphology,studies clearly demonstrate that there exist hemangio- incorporation of DiI-Ac-LDL, and production of hemato-blasts, the common precursors of both hematopoietic poietic cells. We used the LO cells as immunogens toand endothelial cells, in fish, avian, and mammalian em- raise monoclonal antibodies against cell surface anti-bryos, and that VEGF-R2/Flk1, TGFb1, and SCL/Tal-1 gens on LO cells. Among the antibodies obtained, theare essential for the development of hemangioblasts. antibody designated 10B9 (rat IgG1) exhibited very clearHowever, the nature of hemangioblasts remains unex- staining of LO cells but not of NIH3T3 cells by flowplored; in particular, there is as of yet no absolute evi- cytometry (data not shown). This antibody also staineddence that LTR-HSCs are derived from hemangioblasts. the endothelial-like cells in the AGM culture (data notIt also remains uncertain whether hemangioblasts in the shown).yolk sac and the AGM region possess the same differen- Using a standard expression cloning strategy withtiation potential. COS7 cells and 10B9 monoclonal antibody, we isolated

We have attempted to characterize the nature of mam- a cDNA clone of 1.9 kb encoding the 10B9 antigen (Fig-malian hemangioblasts by using our AGM primary cul- ure 2A). DNA sequencing revealed that the C-terminalture system, in which multipotential hematopoietic pro- amino acid sequence including the transmembrane re-genitor cells and endothelial-like cells expand in vitro gion was highly homologous to those of human and(Mukouyama et al., 1998). In this culture, oncostatin M rabbit podocalyxin-like protein 1 (PCLP1) (Kershaw et(OSM), a member of the IL-6 family of cytokines, is es- al., 1995, 1997), suggesting that it is a mouse counterpartsential for the expansion of both cell types. Although of PCLP1. The avian PCLP1 homolog, thrombomucin,OSM plays an essential role in the expansion of both also shares the conserved regions (McNagny et al.,cell populations, OSM does not directly stimulate the 1997). COS7 cells transfected with the PCLP1 cDNAgrowth of hematopoietic progenitors in the colony form- exhibited specific staining with 10B9 antibody (data noting assays (data not shown). We thus hypothesized that shown), confirming that the 10B9 antibody recognizesthe endothelial-like cells may contain hemangioblasts mouse PCLP1. As previously reported (Kershaw et al.,that produce hematopoietic progenitors in the AGM cul- 1997), the amino acid sequence of the N-terminal regionture. In this study, we molecularly cloned a novel marker of PCLP1 is poorly conserved among species. Interest-for endothelial-like cells in the AGM culture and provide ingly, a recent report suggested that both PCLP1 andevidence that LTR-HSCs are generated in vivo from CD34 are ligands for L-selectin in the high endothelialthese hemangioblastic cells in the AGM region. In addi- venule and that PCLP1 and CD34 share common aminotion, we show that OSM can expand angioblasts in vitro. acid sequences in their cytoplasmic tails (Sassetti et al.,

1998). These homologous amino acid residues are alsofound in mouse PCLP1 (broken lines in Figure 2A).Results

PCLP1 was originally identified as a major componentof podocytes in the rabbit kidney and is expressed inGeneration of Hematopoietic Cells

in the AGM Culture some endothelial cells (Kershaw et al., 1995). Consistentwith previous reports (Kershaw et al., 1995, 1997), weIn previous studies using the in vitro culture for AGM

cells, we suggested that the endothelial-like cells in the detected PCLP1 mRNA in kidney, heart, lung, brain, and

PCLP1 as a Marker for Hemangioblasts569

Figure 1. Generation of Hematopoietic Cellsfrom the Endothelial-like Cells in the AGMCulture

The endothelial-like cells generated in the day6 AGM primary culture were pulse labeledwith DiI-Ac-LDL for 6 hr. DiI1CD452 cells (R1fraction) were then isolated by FACS and in-noculated into the unlabeled day 6 AGM cul-ture. After 4 days of chase culture, both float-ing and adherent cells were analyzed byFACS. Newly produced DiI1CD451 cells areshown by brackets.

muscle, but not in spleen, thymus, small intestine, or computerized database maintained at The Jackson Lab-oratory, Bar Harbor, ME). Pclp1 mapped in a region ofliver of adult mice (data not shown). Thrombomucin was

reported to be expressed in thrombocytes and multipo- the composite map that lacks mouse mutations givingthe predictable phenotype for an alteration in this locus.tent hematopoietic progenitors (McNagny et al., 1997).

Likewise, expression of PCLP1 was found in some bone The proximal region of mouse chromosome 6 shares aregion of homology with human chromosome 7q (sum-marrow cells (T. H., unpublished data) and hematopoi-

etic cells in the AGM region (Figure 4A). marized in Figure 2B), which is consistent with the as-signment of the human PCLP1 gene to 7q23-q33 (Ker-shaw et al., 1997). A functionally related gene, CD34,Chromosomal Mapping of the Mouse PCLP1 Genehas been mapped in mouse chromosome 1 (Taketo etThe mouse chromosomal location of the PCLP1 geneal., 1992), and there is no apparent genetic relationship(Pclp1) was determined by interspecific backcross anal-between Pclp1 and CD34 loci.ysis. The mapping results (Figure 2B) indicated that

Pclp1 is located in the proximal region of mouse chro-mosome 6 linked to Met, Cpa, and Ptn (Siracusa et al., Expression of PCLP1 on the Endothelial-like Cell

Clusters in the AGM Culture1991; Ho et al., 1999). The ratios of the total number ofmice exhibiting recombinant chromosomes to the total We examined the expression of PCLP1 on the endothe-

lial-like cells in the AGM culture by immunostaining withnumber of mice analyzed for each pair of loci and themost likely gene order are: centromere - Met - 5/188 - 10B9 anti-PCLP1 antibody. As we expected, PCLP1 was

detectable on endothelial-like cells but not on fibroblas-Cpa - 1/184 - Pcpl1- 10/179 - Ptn. The recombinationfrequencies (expressed as genetic distances in centi- tic cells (data not shown) in the AGM culture. The endo-

thelial-like cells, defined by their polygonal cell morphol-Morgans [cM] 6 the standard error) are - Met - 2.7 61.2 - Cpa - 0.5 6 0.5 - Pclp1 - 5.6 6 1.7 - Ptn. ogy and incorporation of DiI-Ac-LDL, were further

fractionated by fluorescent activated cell sorting (FACS)We have compared our interspecific map of chromo-some 6 with a composite mouse linkage map that using anti-PCLP1 and anti-CD45 antibodies. Except for

the erythroid lineage, CD45 is known to be a panspecificreports the map location of many uncloned mousemutations (provided from Mouse Genome Database, a marker for hematopoietic cells including LTR-HSCs

Immunity570

Figure 2. Primary Sequence and Chromo-some Mapping of Mouse PCLP1

(A) Amino acid sequence of mouse PCLP1.The underline and double underline representsignal peptide and transmembrane domain,respectively. Intracellular regions that areconserved between CD34 and PCLP1 areshown by broken lines.(B) Chromosomal location of mouse PCLP1gene (Pclp1). The segregation patterns ofPclp1 and flanking genes in 173 backcrossanimals that were typed for all loci are shownat the top of the figure. Each column repre-sents the chromosome identified in the back-cross progeny that was inherited from the(C57BL/6J 3 M. spretus) F1 parent. The filledboxes represent the presence of a C57BL/6Jallele, and white boxes represent the pres-ence of an M. spretus allele. The number ofoffspring inheriting each type of chromosomeis listed at the bottom of each column. A par-tial chromosome 6 linkage map showing thelocation of Pclp1 in relation to linked genesis shown at the bottom of the figure. Recom-bination distances between loci in centi-morgans are shown to the left of the chromo-some, and the positions of loci in humanchromosomes, when known, are shown tothe right.

(Morrison et al., 1995). As was consistent with data in indicating that these cells are endothelial-like cells. ItFigure 1, the PCLP11CD452 nonhematopoietic fraction was recently reported that hematogenic angioblasts inin the AGM culture gave rise to CD451 hematopoietic the yolk sac and the P-Sp region express Flk1, VECad-cells in vitro (data not shown). Thus, PCLP1 proved to herin, and CD34 (Nishikawa et al., 1998b). Thus, webe a useful marker to separate hemangioblastic cells in examined whether these molecules are expressed inthe AGM region. the PCLP11CD452 fraction. Interestingly, a majority of

nonhematopoietic CD341 cells, CD311 cells, and Flk11

cells also expressed PCLP1 (Figures 4B and 4C),Expression and Localization of PCLP1 in the AGMwhereas most VECadherin1CD452 cells were found in theRegion of a Mouse EmbryoPCLP12 fraction (data not shown). Consistent with theWe next examined the presence of the PCLP11CD452overlapping expression patterns of PCLP1 and CD34,cells in the intact AGM region of mouse embryos at 11.5expression of these two proteins was colocalized in thedpc. By immunohistochemical staining of transverseendothelium of the dorsal aorta (Figures 3E and 3H) andsections of mouse embryos, PCLP was expressed inin the genital ridge region (Figures 3F and 3I) of thethe AGM region (Figures 3A–3F). When we analyzedmouse embryo at 11.5 dpc. Specific expression ofdissociated cells in the AGM region by FACS staining,PCLP1 in the AGM region was confirmed by in situ hy-there were more PCLP11CD452 cells (32%) than PCLP11

bridization (data not shown). Since the endothelial-likeCD451 cells (1.5%) in the AGM region (Figure 4A). Thecells in the day 6 AGM culture do not express VECad-sorted PCLP11CD452 cells were adherent cells with the

capacity to incorporate DiI-Ac-LDL (data not shown), herin and CD34 (data not shown), we employed PCLP1

PCLP1 as a Marker for Hemangioblasts571

Figure 3. Expression of PCLP1 in the AGM Region of a Mouse Embryo

Immunohistochemical staining of paraffin sections of the AGM region of a mouse embryo at 11.5 dpc with isotype control (A–C), anti-PCLP1(D–F), or anti-CD34 (G–I) antibodies. Specific expression of PCLP1 or CD34 is shown by the brown color. Original magnification was 403 (leftpanels). Aorta and genital ridge regions are further enlarged in middle and right panels, respectively.

as a marker for the separation of hemangioblasts in the of OSM, VEGF, and bFGF. The resultant cells expressedhigher levels of CD34 and VECadherin (Figure 5B) thanAGM region in this work.those in the initial cell population (Figure 4B; data notshown) or those cultured with OSM alone (Figure 5B).Endothelial Differentiation of the PCLP11CD452

Cells from the AGM Region Coculture of the PCLP11CD452 cells with OP9 also re-sulted in increased expression of CD31 and decreasedThe PCLP11CD452 cell fraction was separated by cell

sorting from the AGM region of mouse embryos at 11.5 expression of PCLP1 (Figure 5B). Moreover, the OP9cocultured cells formed a vascular network on a matrigeldpc. CD451 cells were undetectable in the sorted cells

by reanalysis (data not shown). Even 3 hr after plating plate (Figure 5C), while the cells grown without OP9failed to form a network (data not shown). These resultsof the sorted PCLP11CD452 cells, no hematopoietic-

like cells could be detected by microscopic observation, indicate that PCLP11CD452 cells in the AGM regionare able to differentiate to endothelial cells in the pres-and the cells were capable of incorporating DiI-Ac-LDL

(data not shown). When these cells were cultured in ence of OP9, OSM, VEGF, and bFGF. Therefore, thePCLP11CD452 cells are likely to be the endothelial pre-the presence of OSM for 6 days, endothelial-like cells

increased by 10-fold during incubation (Figure 5A) and all cursor cells, i.e., angioblasts. Growth of the angioblastsin the AGM region appears to be OSM dependent, andadherent cells incorporated DiI-Ac-LDL and expressed

Flk1 (data not shown), whereas no cells grew in the their differentiation requires additional factors includingVEGF, bFGF, and unknown factors produced from OP9absence of OSM. Since only 12% of the sorted PCLP11

CD452 cells were Flk11 cells at the time of separation cells.On the other hand, coculture of PCLP11CD452 cells(Figure 4C), Flk11 cells may be selectively expanded or

Flk1 expression may be induced during cultivation. The with OP9 cells in the presence of hematopoietic growthfactors containing SCF, interleukin(IL)-3, and erythropoi-PCLP11CD452 cells grown in the presence of OSM were

partially positive for CD31, negative for CD34, and very etin (EPO) resulted in the development of hematopoieticcells. The hematopoietic cells included Mac-1/Gr-weakly positive for VECadherin (Figure 5B).

To test the possibility that the PCLP11CD452 cells 1-positive myeloid cells, B220/Thy-1-positive lymphoidcells, and Ter119-positive erythroid cells (data notdifferentiate to endothelial cells, we employed the OP9

coculture system that has been used to induce endothe- shown), suggesting that multiple lineages of hematopoi-etic cells were generated from the PCLP11CD452 cellslial differentiation in vitro (Hirashima et al., 1999). The

PCLP11CD452 cells from the AGM region were cocul- in vitro. Generation of these hematopoietic cells wasalso OSM dependent. Taken together with the data fromtured with OP9 stromal cells for 10 days in the presence

Immunity572

Figure 4. FACS Analyses of the AGM Region of Mouse Embryos

A single cell suspension prepared from the AGM regions of mouse embryos at 11.5 dpc was stained with anti-PCLP1 antibody and anti-CD45antibody (A), anti-CD34 (B), anti-CD31 (B), or anti-Flk1 antibody (C) and analyzed by FACS. Patterns with appropriate isotype controls arealso shown. In (B) and (C), FACS patterns in the CD452 cell fraction are presented.

the DiI-Ac-LDL labeling experiment (Figure 1), we con- injection of the GFP1PCLP11CD452 cells. To repopulatethe donor-derived blood cells, the injection of more thanclude that PCLP11CD452 cells in the AGM region con-

tain hemangioblasts and angioblasts. 1.7 3 105 cells was required, indicating that a smallfraction of the PCLP11CD452 cell population are capa-ble of generating LTR-HSCs. In contrast, no donor-Generation of LTR-HSCs from the PCLP11CD452

Cells in the AGM Region derived hematopoietic cells were found by injecting thesame number of PCLP1-CD452 cells from the AGM re-A major goal of this study was to find out whether hem-

angioblasts in the AGM region can give rise to LTR- gion (Table 1). Moreover, injection of the PCLP11CD451

cells (0.9–2.1 3 104) did not contribute to GFP chime-HSCs in vivo. LTR-HSCs were detected among the he-matopoietic progenitors expanded in the AGM culture rism, which effectively excluded the possibility that a

small number of contaminating CD451 cells in theby the standard repopulation assay using irradiatedadult mice. However, they were more efficiently en- PCLP11CD452 fraction repopulated the recipient mice.

The chimerism was maintained up to 6 months in bothgrafted when injected into livers of busulfan-treatedneonatal mice (M. Takeuchi et al., unpublished data). myeloid and lymphoid compartments of the peripheral

blood of the mouse injected with 1.7 3 105 of theThis is reasonable, as LTR-HSCs generated in the AGMregion seed the fetal liver in vivo before homing into the PCLP11CD452 cells (Figure 6). In this repopulated

mouse, all lineages of donor-derived GFP-positive he-bone marrow. We thus considered the possibility thatif LTR-HSCs are generated from the hemangioblasts matopoietic cells were repopulated in the spleen and

bone marrow (data not shown). CD4/CD8 double-posi-present in the AGM region, they would engraft the neo-natal liver more efficiently than in irradiated adult mice. tive and mature single-positive T cells derived from the

donor were also detected in the recipient thymus (dataAccording to the recently established procedure (Yoderet al., 1997), we injected the PCLP11CD452 cells from not shown). These results indicate that hemangioblasts

with potential to generate LTR-HSCs are present in thethe AGM regions of green fluorescent protein (GFP)transgenic mouse embryos at 11.5 dpc into the busul- PCLP11CD452 cell population of the AGM region. A

similar number of the PCLP11CD452 cells derived fromfan-treated nontransgenic neonatal mice.As summarized in Table 1, donor-derived GFP-posi- the day 6 AGM primary culture showed a decreased

repopulation potential (Table 1), although they were ca-tive hematopoietic cells were detected in the peripheralblood of seven out of nine mice at 2–6 months after the pable of generating hematopoietic cells in vitro (data

PCLP1 as a Marker for Hemangioblasts573

Figure 5. Differentiation of PCLP11CD452 Cells into Endothelial Cells

(A) Morphological appearance of the sorted PCLP11CD452 cells from the AGM region after 6 days in culture with SCF and OSM.(B) Induction of various endothelial cell marker expression in the PCLP11CD452 cells after coculture with OP9 stromal cells for 10 days inthe presence of OSM, VEGF, and bFGF (right panels). OP9 cells were gated out by the forward scatter window. In left panels, FACS patternsof the PCLP11CD452 cells after 6 days in culture in the absence of OP9 are shown. Thick and thin lines show staining patterns of isotypecontrol and indicated antibodies, respectively.(C) Vascular network formation of the endothelial cells. After 10 days in coculture of the PCLP11CD452 cells with OP9, cells were placed onthe matrigel and cultured for 12 hr.

not shown). Hence, the in vitro culture of the AGM cells PCLP11CD452 cell population of the mouse AGM re-gion. Our data also indicate that the PCLP11CD452 cellsnegatively affected the repopulation potential of the

PCLP11CD452 cells. possess angioblastic activity and that OSM is essentialfor their growth and differentiation. On the other hand,the DiI1CD452 endothelial-like cells in the AGM primaryDiscussionculture generated hematopoietic cells in vitro (Figure 1),and the sorted PCLP11CD452 cells also gave rise toAlthough it has been known for some time that heman-

gioblasts are the precursors of both hematopoietic and hematopoietic cells in vitro in the presence of hemato-poietic growth factors and OP9 cells. Most importantly,endothelial cells, the precise nature of the hemangio-

blasts has remained obscure. In this study, we pre- PCLP11CD452 cells were shown to reconstitute the en-tire and long-term hematopoietic system when trans-sent evidence that hemangioblasts are present in the

Table 1. Hematopoietic Cell Generation from PCLP11CD452 Cells in Injected Mice

Experiment Source Fraction Cell Number/Mouse Engrafted/Total (% Chimerism)

1 AGM PCLP11CD452 1.7 3 105 1/1* (53%)PCLP11CD451 9.0 3 103 0/1*

2 AGM PCLP11CD452 2.8 3 105 2/2 (21%, 1.2%)PCLP12CD452 2.8 3 105 0/3

3 AGM PCLP11CD452 3.4 3 105 1/1 (29%)PCLP12CD452 3.4 3 105 0/1

4 AGM PCLP11CD452 2.2 3 105 3/5 (4.4%, 2.6%, 1.5%)PCLP11CD451 2.1 3 104 0/4

5 AGM culture PCLP11CD452 3.0 3 105 2/5 (0.35%, 0.23%)PCLP11CD451 2.9 3 104 0/2

6 AGM culture PCLP11CD452 3.0 3 105 2/5 (0.14%, 0.14%)PCLP12CD452 3.0 3 105 0/2

Each cell fraction was sorted from AGM region or AGM culture of GFP mice and injected into busulfan-treated neonatal mice. Peripheralblood was taken at 2 months (*6 months) after injection and subjected to FACS analysis. Relative frequency of GFP1 cells in engrafted micewas calculated and expressed as percentage of chimerism.

Immunity574

Figure 6. Induction of LTR-HSCs In Vivo fromthe PCLP11CD452 Hemangioblasts

PCLP11CD452 or PCLP11CD451 cells wereisolated from the AGM region of GFPtransgenic mice and injected into the liver ofneonatal C57BL/6 mice at 36 hr after birth.After 6 months, peripheral blood was takenfrom the PCLP11CD452 injected mouse, andcontrol mice and mononuclear cells werestained with various hematopoietic lineageantibodies as indicated. FACS patterns ofC57BL/6 and GFP mouse are shown as nega-tive and positive controls, respectively.

planted into busulfan-treated neonatal mice (Figure 6; In this study, we have identified PCLP1 as a marker forhemangioblasts. PCLP1 is a highly glycosylated proteinTable 1). As more than 1.7 3 105 PCLP11CD452 cells

were required for long-term reconstitution of the entire with some similarity to CD34, a conventional markerfor LTR-HSCs. Interestingly, both PCLP1 and CD34 arehematopoietic system, LTR-HSCs must be generated

from a minor fraction of the PCLP11CD452 cells. At ligands for L-selectin in the lymphocyte-high endothe-lium venule and contain conserved amino acid se-present, it remains uncertain whether LTR-HSCs and

angioblasts are derived from the same precursor, i.e., quences in their cytoplasmic regions (Sassetti et al.,1998), suggesting an overlapping function betweenhemangioblasts, or their precursors are different. Be-

cause some hematopoietic cells are directly derived the two molecules. In fact, both PCLP1 and CD34 areexpressed in the dorsal aortic endothelium and thefrom the PCLP11CD452 endothelial-like cells that incor-

porate DiI-Ac-LDL, the precursors of LTR-HSCs would genital ridge region (Figures 3D–3I), and as much as91% of CD341 cells in the AGM region also expressexhibit characteristics very similar to endothelial cells

even if the origins of HSCs and angioblasts are distinct. PCLP1 (Figure 4B). In chicken, thrombomucin, an aviancounterpart of PCLP1, is expressed in hematopoieticWhile endothelial cells and hematopoietic cells are most

likely to be derived from the common precursors, i.e., progenitors and thrombocytes (McNagny et al., 1997);however, no avian CD34 has been identified. Therefore,hemangioblasts, the final proof of the model awaits puri-

fication of the hemagioblast and demonstration of its the functions, if any, of CD34 might be compensatedby PCLP1. Sanchez et al. (1996) showed that LTR-HSCspotential to generate both lineages. Since the repopula-

tion efficiency of the PCLP11CD452 cells in the AGM in the AGM region are CD341c-Kit1, and LTR-HSCsin the yolk sac and P-Sp region were also found inculture was significantly lower than that of the AGM

region itself (Table 1), LTR activity of the hemangioblasts the CD341c-Kit1 fraction (Yoder et al., 1997). Sincethere exist PCLP11CD341CD452 cells (12% of thecould be lost during in vitro cultivation. Alternatively, the

angioblasts might be preferentially amplified by OSM. PCLP11CD452 cells) in the AGM region (Figure 4B), theLTR-HSC activity found in the CD341 fraction might rep-Here again, purification of the AGM hemangioblasts and

cell fate tracing experiments will be necessary to distin- resent, in part, that of the hemangioblasts. The repopula-tion capacity of the PCLP11CD341CD452 cells in theguish between these possibilities.

PCLP1 as a Marker for Hemangioblasts575

AGM region is currently under investigation. As the dor- activity in the Flk11VECadherin1CD452 population de-rived from the yolk sac and in the P-Sp region of mousesal aortic endothelium has been suggested to containembryos at 9.5 dpc. Studies of in vitro ES differentiationthe budding sites of hematopoietic cells (Tavian et al.,and Flk1 knockout mice also suggested that putative1996, 1999), hemangioblasts are likely to be present inhemangioblasts express Flk1 (Shalaby et al., 1995; Choithat region. However, our study also raises the possibil-et al., 1998; Nishikawa et al., 1998a). Consistent withity that hemangioblasts might also exist in the genitalthese results, a fraction (12%) of the PCLP11CD452 cellsridge region (Figures 3F and 3I).in the AGM region expressed Flk1 (Figure 4C), and theGene targeting experiments have revealed that a num-cultured PCLP11CD452 cells in the presence of OSMber of genes are critically involved in definitive hemato-were all Flk1 positive (data not shown). In contrast, thepoiesis. However, disruption of such genes often leadsfrequency of VECadherin1 cells in the PCLP11CD452to embryonic death, which makes functional analysis ofpopulation in the AGM region was very low (3%) (datathe gene difficult. Our in vitro culture system of the AGMnot shown). Although hematopoietic cells such as my-cells provides a convenient way to analyze the functioneloid, lymphoid, and erythroid cells were shown to beof such genes. Hematopoietic transcription factors c-Mybderived from hemangioblasts of the P-Sp, yolk sac, orand AML-1 are known to be essential for definitive hema-ES cells, none of these cells have been shown to gener-topoiesis, and disruption of either one of these genesate LTR-HSCs (Nishikawa et al., 1998a, 1998b), raisingresults in embryonic lethality due to the lack of definitivethe possibility that these hemangioblasts are distincthematopoiesis (Mucenski et al., 1991; Okuda et al., 1996).from those in the AGM region. As production and expan-In the AGM culture of the c-Myb knockout mice, no hema-sion of LTR-HSCs from ES cells is an important chal-topoietic cells were produced, whereas endothelial-likelenge for future applications of human ES cells, it wouldcells were generated normally. Moreover, retrovirus-medi-be interesting to test if PCLP11CD452 cells derived fromated expression of c-Myb in the c-Myb-deficient AGMthe ES cells are able to generate LTR-HSCs. It alsoculture restored the production of hematopoietic cellsremains unknown whether hemangioblasts exist in other(Mukouyama et al., 1999). These results indicate thattissues. CD341 circulating endothelial precursors werec-Myb is required for the development of hematopoieticrecently found in the peripheral blood and were showncells from hemangioblasts but not for the developmentto be derived from the bone marrow (Takahashi et al.,of hemangioblasts/angioblasts. Likewise, no hemato-1999). For future clinical applications of hemagioblasts,poietic cells were produced in the AGM culture of AML-1it will be important to search for hemangioblasts in new-knockout mice, while endothelial-like cells were nor-born and adult tissues such as umbilical cord and bonemally produced, suggesting an essential role for AML-1marrow.in the development of LTR-HSCs from hemangioblasts

(Y. M. et al., submitted). In contrast to definitive hemato-Experimental Procedurespoiesis, primitive hematopoiesis in the yolk sac was

normal in both c-Myb and AML-1 knockout mice, indi- Micecating a distinct mechanism of hematogenesis in the Timed pregnant C57BL/6 mice were purchased from Nihon SLC

(Hamamatsu, Japan). The GFP transgenic mice (Okabe et al., 1997)yolk sac. Therefore, the question arises whether thewere maintained and mated in our animal facility. The time at middayhemagioblasts in the yolk sac, if present, are the same(12:00) was taken to be 0.5 dpc for the plugged mice.as those in the AGM region.

Like the AGM cultures derived from embyros at 11.5 Cell Labeling Experimentdpc, OSM-dependent expansion of the hematopoietic AGM cells in the day 6 culture (Mukouyama et al., 1998) were exten-progenitors and the endothelial-like cells was observed sively washed with medium to remove hematopoietic cells and incu-

bated with 10 mg/ml acetylated low-density lipoprotein labeledin the primary culture of the P-Sp region of embryos atwith 1,19-dioctadecyl-3,3,39,39-tetramethylindo-carbocyanine per-9.5 dpc. However, no such expansion was observed inchlorate (DiI-Ac-LDL) (Biomedical Technologies) for 6 hr at 378C.the yolk sac of embryos at 9.5 dpc. Nevertheless, theAfter washing with PBS twice, the AGM cells were subjected to

PCLP11CD452 cells were present in both the yolk sac staining with fluorescein isothiocyanate (FITC)-conjugated anti-and the P-Sp region (T. H., unpublished data). It remains CD45 antibody (PharMingen).to be determined whether the PCLP11CD452 cells in

Generation of Monoclonal Antibodiesthe yolk sac can generate LTR-HSCs. Also, the possibil-Wistar rats (Nihon SLC) were immunized with 107 of LO cells in theity that lack of an OSM response by PCLP11CD452 cellspresence of Freund’s complete adjuvant (WAKO, Osaka, Japan) byin the yolk sac is due to OSM receptor deficiency inthe standard immunization procedure (Hockfield et al., 1993). Lymph

the yolk sac should be investigated. In contrast to the nodes were dissociated and fused with mouse myeloma P3X cellsabsolute requirement of Flk1 for both hematopoiesis by polyethylene glycol as previously described (Ogorochi et al.,

1992), and hybridoma supernatants were screened for the produc-and vasculogenesis (Shalaby et al., 1995, 1997), thetion of anti-LO-specific antibodies by FACS. 10B9 monoclonal anti-OSM/OSMR system does not seem to be indispensablebody was chosen based on the specific staining of the endothelial-because gp130 knockout mice exhibited a less severelike cells in the AGM culture. 10B9 antibody was produced in nudedefect in hematopoiesis and retained an almost normal mice and purified by using E-Z-Sep (Pharmacia Biotech, Uppsala,

vasculature system compared to Flk1-deficient mice Sweden). The isotype of the 10B9 antibody was determined by using(Yoshida et al., 1996). This suggests that an OSM com- the rat IgG isotyping kit (Serotec, Oxford, UK). Biotinylated 10B9

antibody was prepared by using Enzotin (Enzo Diagnostics) ac-pensatory mechanism exists to stimulate hemangio-cording to the manufacturer’s instruction.blasts/angioblasts in gp130 knockout mice. By using

the in vitro culture system for hemangioblasts/angio-Expression Cloning of PCLP1 cDNA, Sequencing,

blasts, the nature of additional cytokines needed for and Northern Blot Analysistheir expansion can be explored. Expression cloning of a cDNA encoding the 10B9 antigen was car-

ried out by using COS7 cells as previously described (Harada etNishikawa et al. (1998b) recently showed hematogenic

Immunity576

al., 1990) except that magnetic beads conjugated with anti-rat IgG For matrigel assays, cells (2 3 105) were resuspended in Dulbec-co’s modified Eagle’s medium containing 1% FCS and VEGF (10(Dynabeads M-450) (Dynal, Oslo, Norway) were employed instead

of plate panning. ng/ml) and overlayed on a Biocoat matrigel basement membrane(Becton Dickinson) in a 6-well plate.The DNA sequences of the cDNAs were determined by using a

Dye terminator cycle sequencing kit (Perkin Elmer) and an auto-mated DNA sequencer (Applied Biosystems). Northern blot analysis

Transplantation Assaywas carried out as previously described (Tanaka et al., 1999) byTransplantation of cells into busulfan-treated neonatal mice wasusing digoxigenin (DIG)-labeled single strand DNA probe for theperformed as previously described (Yoder et al., 1997) with a slightPCLP1 cDNA (2.1 kb) as a probe.modification. In brief, pregnant C57BL/6 mice were intraperitoniallyinjected with busulfan (Sigma) at 12.5 mg/g on pregnant day 17 and18. Within 24–48 hr after birth, cells derived from GFP mice in 25Interspecific Mouse Backcross Mapping

Interspecific backcross progeny were generated by mating (C57BL/ ml of PBS were injected into the liver.6J 3 Mus spretus) F1 females and C57BL/6J males as described(Copeland and Jenkins, 1991). A total of 205 N2 mice were used to

Acknowledgmentsmap the Pclp1 locus by Southern blot hybridization by using anz1.5 kb EcoRI/NotI fragment as a probe. Fragments of 5.6 and 5.1

We are grateful to S. Nishikawa (Kyoto University) for valuable sug-kb were detected in ScaI-digested C57BL/6J DNA, and fragmentsgestions and for providing us with anti-VECadherin antibody andof 12.0 and 5.1 kb were detected in ScaI-digested M. spretus DNA.OP9 cell line. We thank M. Yoder (Indiana University) for kind instruc-tions on neonatal liver injection and Debra J. Gilbert for excellenttechnical assistance during chromosome mapping. M. T. and Y. M.Flow Cytometry and Cell Sorting

Isolated AGM regions were dissociated by incubation with dispase are recipients of Research Fellowships of the Japan Society for thePromotion of Science for young scientists. This work was partly(Boehringer) for 30 min at 378C and cell dissociation buffer (Gibco-

BRL) for 30 min at 378C followed by vigorous agitation. Cells were supported by Grants-in-Aid for Scientific Research from Monbusho,grants from CREST, The Takeda Foundation, Riken, and The Organi-then stained with a primary antibody followed by one of various

fluorescent-labeled secondary reagents as previously described zation for Pharmaceutical Safety and Research. This research wasalso supported, in part, by the National Cancer Institute, Department(Mukouyama et al., 1998). Cells were finally resuspended in 0.5 ml

of PBS containing propidium iodide (PI) (Sigma) and analyzed by of Health and Human Services, under contract with Advanced Bio-science Laboratories.FACS Calibur (Becton Dickinson). PI-positive dead cells were ex-

cluded. The commercially available monoclonal antibodies used forFACS were described previously (Mukouyama et al., 1999). Anti-

Received September 16, 1999; revised October 11, 1999.VECadherin antibody (VECD1) (Matsuyoshi et al., 1997) was kindlyprovided by S. Nishikawa (Kyoto University).

For cell sorting, AGM regions from GFP-positive embryos at 11.5 Referencesdpc were trypsinized as described above, and cells (107/ml) wereincubated with biotinylated anti-PCLP1 antibody at 10 mg/ml in 5% Choi, K., Kennedy, M., Kazarov, A., Papadimitriou, J.C., and Keller,fetal calf serum (FCS)-PBS on ice for 30 min. After washing with 20- G. (1998). A common precursor for hematopoietic and endothelialfold volumes of 5% FCS-PBS, cells were stained with phycoerythrin cells. Development 125, 725–732.(PE)-conjugated anti-CD45 antibody (10 mg/ml) and allophycocyanin

Copeland, N.G., and Jenkins, N.A. (1991). Development and applica-(APC)-conjugated streptavidin (Molecular Probe) (10 mg/ml) on icetions of a molecular genetic linkage map of the mouse genome.for 30 min and subjected to cell sorting using FACS Vantage (BectonTrends. Genet. 7, 113–118.Dickinson). In a typical case, out of 1.1 3 107 cells obtained fromDelassus, S., and Cumano, A. (1996). Circulation of hematopoietic40 AGM regions, 8.5 3 105 of PCLP11CD452 cells, 1.0 3 106 ofprogenitors in the mouse embryo. Immunity 4, 97–106.PCLP1-CD452 cells, and 5.9 3 104 of PCLP11CD451 cells were

obtained by cell sorting. Dickson, M.C., Martin, J.S., Cousins, F.M., Kulkarni, A.B., Karlsson,S., and Akhurst, R.J. (1995). Defective haematopoiesis and vasculo-genesis in transforming growth factor-beta 1 knock out mice. Devel-

Immunohistochemical Staining and In Situ Hybridization opment 121, 1845–1854.For preparation of paraffin sections, the caudal half of mouse em-

Dzierzak, E., Medvinsky, A., and de Bruijn, M. (1998). Qualitativebryos at 11.5 dpc was fixed in 4% paraformaldehyde (PFA)-PBS forand quantitative aspects of haematopoietic cell development in the10 hr. Paraffin sections (6 mm thick) were hydrated and stained withmammalian embryo. Immunol. Today 19, 228–236.anti-PCLP1 or anti-CD34 antibody at 10 mg/ml at 48C overnight.Eichmann, A., Corbel, C., Nataf, V., Vaigot, P., Breant, C., and LeAfter incubation with peroxidase-conjugated anti-rat IgG (Amer-Douarin, N.M. (1997). Ligand-dependent development of the endo-sham), signals were visualized by 3,39-diaminobenzidine (DAB) asthelial and hemopoietic lineages from embryonic mesodermal cellspreviously described (Hara et al., 1998). Samples were counter-expressing vascular endothelial growth factor receptor 2. Proc. Natl.stained with methylgreen. In situ hybridization of the paraffin sec-Acad. Sci. USA 94, 5141–5146.tions was carried out by using the 59-part probe of the PCLP1 cDNA

fragment (nucleotide 126–354) as previously described (Imakawa et Gering, M., Rodaway, A.R., Gottgens, B., Patient, R.K., and Green,A.R. (1998). The SCL gene specifies haemangioblast developmental., 1995).from early mesoderm. EMBO J. 17, 4029–4045.

Godin, I., Dieterlen-Lievre, F., and Cumano, A. (1995). EmergenceOP9 Coculture Experiment and Matrigel Assayof multipotent hemopoietic cells in the yolk sac and paraaorticMouse calvaria-derived OP9 cells (provided by S. Nishikawa) weresplanchnopleura in mouse embryos, beginning at 8.5 days postco-maintained as previously described (Kodama et al., 1994). Sorteditus. Proc. Natl. Acad. Sci. USA 92, 773–777.PCLP11CD452 cells from the AGM region were innoculated on sub-Hara, T., Tamura, K., de Miguel, M.P., Mukouyama, Y., Kim, H.,confluent OP9 cells in 6-well plates with the AGM culture mediumKogo, H., Donovan, P.J., and Miyajima, A. (1998). Distinct roles ofcontaining various cytokines and cultured for 10 days. For the gener-oncostatin M and leukemia inhibitory factor in the developmentation of hematopoietic cells, 5 3 104 cells were cocultured with OP9of primordial germ cells and sertoli cells in mice. Dev. Biol. 201,in the presence of SCF (kind gift from Kirin Brewery, Takasaki, Japan)144–153.(100 ng/ml), bFGF (1 ng/ml), LIF (10 ng/ml), OSM (10 ng/ml), IL-3

(10 ng/ml), and EPO (gift from Kirin) (2 U/ml). For endothelial cell Harada, N., Castle, B.E., Gorman, D.M., Itoh, N., Schreurs, J., Barrett,R.L., Howard, M., and Miyajima, A. (1990). Expression cloning of adifferentiation, 104 cells were cocultured in the presence of OSM

(10 ng/ml), bFGF (1 ng/ml), and VEGF (PeproTech, London, UK) (10 cDNA encoding the murine interleukin 4 receptor based on ligandbinding. Proc. Natl. Acad. Sci. USA. 87, 857–861.ng/ml).

PCLP1 as a Marker for Hemangioblasts577

Hirashima, M., Kataoka, H., Nishikawa, S., and Matsuyoshi, N. Ogorochi, T., Hara, T., Wang, H.M., Maruyama, K., and Miyajima, A.(1999). Maturation of embryonic stem cells into endothelial cells in (1992). Monoclonal antibodies specific for low-affinity interleukin-3an in vitro model of vasculogenesis. Blood 93, 1253–1263. (IL-3) binding protein AIC2A: evidence that AIC2A is a component

of a high-affinity IL-3 receptor. Blood 79, 895–903.Ho, H.T., Jenkins, N.A., Copeland, N.G., Gilbert, D.J., Winkles, J.A.,Louie, H.W., Lee, F.K., Chung, S.S., and Chung, S.K. (1999). Compar- Okabe, M., Ikawa, M., Kominami, K., Nakanishi, T., and Nishimune,isons of genomic structures and chromosomal locations of the Y. (1997). “Green mice” as a source of ubiquitous green cells. FEBSmouse aldose reductase and aldose reductase-like genes. Eur. J. Lett. 407, 313–319.Biochem. 259, 726–730. Okuda, T., van Deursen, J., Hiebert, S.W., Grosveld, G., and Down-Hockfield, S., Carlson, S., Evans, C., Levitt, P., Pintar, J., and Sib- ing, J.R. (1996). AML1, the target of multiple chromosomal transloca-erstein, L. (1993). Selected Methods for Antibody and Nucleic Acid tions in human leukemia, is essential for normal fetal liver hemato-Probes, Volume 1 (New York: Cold Spring Harbor Laboratory Press). poiesis. Cell 84, 321–330.Imakawa, K., Tamura, K., McGuire, W.J., Khan, S., Harbison, L.A., Pardanaud, L., and Dieterlen-Lievre, F. (1999). Manipulation of theStanga, J.P., Helmer, S.D., and Christenson, R.K. (1995). Effect of angiopoietic/hemangiopoietic commitment in the avian embryo. De-interleukin-3 on ovine trophoblast interferon during early conceptus velopment 126, 617–627.development. Endocrine 3, 511–517. Pardanaud, L., Luton, D., Prigent, M., Bourcheix, L.M., Catala, M.,Jaffredo, T., Gautier, R., Eichmann, A., and Dieterlen-Lievre, F. and Dieterlen-Lievre, F. (1996). Two distinct endothelial lineages in(1998). Intraaortic hemopoietic cells are derived from endothelial ontogeny, one of them related to hemopoiesis. Development 122,cells during ontogeny. Development 125, 4575–4583. 1363–1371.Kershaw, D.B., Beck, S.G., Wharram, B.L., Wiggins, J.E., Goyal, Porcher, C., Swat, W., Rockwell, K., Fujiwara, Y., Alt, F.W., and Orkin,M., Thomas, P.E., and Wiggins, R.C. (1997). Molecular cloning and S.H. (1996). The T cell leukemia oncoprotein SCL/tal-1 is essentialcharacterization of human podocalyxin-like protein. Orthologous re- for development of all hematopoietic lineages. Cell 86, 47–57.lationship to rabbit PCLP1 and rat podocalyxin. J. Biol. Chem. 272, Sabin, F.R. (1920). Studies on the origin of blood vessels and of red15708–15714.

corpuscles as seen in the living blastoderm of the chick during theKershaw, D.B., Thomas, P.E., Wharram, B.L., Goyal, M., Wiggins, second day of incubation. Contrib. Embryol. 9, 213–262.J.E., Whiteside, C.I., and Wiggins, R.C. (1995). Molecular cloning,

Sanchez, M.J., Holmes, A., Miles, C., and Dzierzak, E. (1996). Char-expression, and characterization of podocalyxin-like protein 1 from

acterization of the first definitive hematopoietic stem cells in therabbit as a transmembrane protein of glomerular podocytes and

AGM and liver of the mouse embryo. Immunity 5, 513–525.vascular endothelium. J. Biol. Chem. 270, 29439–29446.

Sassetti, C., Tangemann, K., Singer, M.S., Kershaw, D.B., and Ro-Kodama, H., Nose, M., Niida, S., Nishikawa, S., and Nishikawa, S.-I.

sen, S.D. (1998). Identification of podocalyxin-like protein as a high(1994). Involvement of the c-kit receptor in the adhesion of hemato-

endothelial venule ligand for L-selectin: parallels to CD34. J. Exp.poietic stem cells to stroma cells. Exp. Hematol. 22, 979–984.

Med. 187, 1965–1975.Liao, E.C., Paw, B.H., Oates, A.C., Pratt, S.J., Postlethwait, J.H., and

Schuh, A.C., Faloon, P., Hu, Q.L., Bhimani, M., and Choi, K. (1999).Zon, L.I. (1998). SCL/Tal-1 transcription factor acts downstream ofIn vitro hematopoietic and endothelial potential of flk-1(2/2) embry-cloche to specify hematopoietic and vascular progenitors in zebra-onic stem cells and embryos. Proc. Natl. Acad. Sci. USA 96, 2159–fish. Genes Dev. 12, 621–626.2164.

Matsuyoshi, N., Toda, K., Horiguchi, Y., Tanaka, T., Nakagawa, S.,Shalaby, F., Rossant, J., Yamaguchi, T.P., Gertsenstein, M., Wu,Takeichi, M., and Imamura, S. (1997). In vivo evidence of the criticalX.F., Breitman, M.L., and Schuh, A.C. (1995). Failure of blood-islandrole of cadherin-5 in murine vascular integrity. Proc. Assoc. Am.formation and vasculogenesis in Flk-1-deficient mice. Nature 376,Physicians 109, 362–371.62–66.

McNagny, K.M., Pettersson, I., Rossi, F., Flamme, I., Shevchenko,Shalaby, F., Ho, J., Stanford, W.L., Fischer, K.D., Schuh, A.C.,A., Mann, M., and Graf, T. (1997). Thrombomucin, a novel cell surfaceSchwartz, L., Bernstein, A., and Rossant, J. (1997). A requirementprotein that defines thrombocytes and multipotent hematopoieticfor Flk1 in primitive and definitive hematopoiesis and vasculogen-progenitors. J. Cell Biol. 138, 1395–1407.esis. Cell 89, 981–990.

Morrison, S.J., Uchida, N., and Weissman, I.L. (1995). The biologySiracusa, L.D., Jenkins, N.A., and Copeland, N.G. (1991). Identifica-of hematopoietic stem cells. Annu. Rev. Cell. Dev. Biol. 11, 35–71.tion and applications of repetitive probes for gene mapping in the

Mucenski, M.L., McLain, K., Kier, A.B., Swerdlow, S.H., Schreiner,mouse. Genetics 127, 169–179.

C.M., Miller, T.A., Pietryga, D.W., Scott, W.J., Jr., and Potter, S.S.Takahashi, T., Kalka, C., Masuda, H., Chen, D., Silver, M., Kearney,(1991). A functional c-myb gene is required for normal murine fetalM., Magner, M., Isner, J.M., and Asahara, T. (1999). Ischemia-hepatic hematopoiesis. Cell 65, 677–689.and cytokine-induced mobilization of bone marrow-derived endo-

Mukouyama, Y., Hara, T., Xu, M., Tamura, K., Donovan, P.J., Kim,thelial progenitor cells for neovascularization. Nat. Med. 5,

H., Kogo, H., Tsuji, K., Nakahata, T., and Miyajima, A. (1998). In vitro434–438.expansion of murine multipotential hematopoietic progenitors fromTaketo, M., Howard, T.A., and Seldin, M.F. (1992). Mapping of re-the embryonic aorta-gonad-mesonephros region. Immunity 8,combinant retrovirus integration sites that cause expression of the105–114.viral genome in murine embryonal carcinoma cells. Mamm. GenomeMukouyama, Y., Chiba, N., Mucenski, M.L., Satake, M., Miyajima,2, 240–245.A., Hara, T., and Watanabe, T. (1999). Hematopoietic cells in culturesTanaka, M., Hara, T., Copeland, N.G., Gilbert, D.J., Jenkins, N.A.,of the murine embryonic aorta-gonad-mesonephros region are in-and Miyajima, A. (1999). Reconstitution of the functional mouseduced by c-Myb. Curr. Biol. 9, 833–836.oncostatin M (OSM) receptor: molecular cloning of the mouse OSMMurray, P.D.F. (1932). The development in vitro of the blood of thereceptor beta subunit. Blood 93, 804–815.early chick embryo. Proc. Roy. Soc. London 11, 497–521.Tavian, M., Coulombel, L., Luton, D., Clemente, H.S., Dieterlen-Lievre,Nishikawa, S.I., Nishikawa, S., Hirashima, M., Matsuyoshi, N., andF., and Peault, B. (1996). Aorta-associated CD341 hematopoieticKodama, H. (1998a). Progressive lineage analysis by cell sorting andcells in the early human embryo. Blood 87, 67–72.culture identifies FLK11VE-cadherin1 cells at a diverging point ofTavian, M., Hallais, M.F., and Peault, B. (1999). Emergence of in-endothelial and hemopoietic lineages. Development 125, 1747–traembryonic hematopoietic precursors in the pre-liver human em-1757.bryo. Development 126, 793–803.Nishikawa, S.I., Nishikawa, S., Kawamoto, H., Yoshida, H., Kizumoto,Visvader, J.E., Fujiwara, Y., and Orkin, S.H. (1998). Unsuspected roleM., Kataoka, H., and Katsura, Y. (1998b). In vitro generation of lympho-for the T-cell leukemia protein SCL/tal-1 in vascular development.hematopoietic cells from endothelial cells purified from murine em-

bryos. Immunity 8, 761–769. Genes Dev. 12, 473–479.

Immunity578

Voyta, J.C., Via, D.P., Butterfield, C.E., and Zetter, B.R. (1984). Identi-fication and isolation of endothelial cells based on their increaseduptake of acetylated-low density lipoprotein. J. Cell Biol. 99, 2034–2040.

Yoder, M.C., Hiatt, K., Dutt, P., Mukherjee, P., Bodine, D.M., andOrlic, D. (1997). Characterization of definitive lymphohematopoieticstem cells in the day 9 murine yolk sac. Immunity 7, 335–344.

Yoshida, K., Taga, T., Saito, M., Suematsu, S., Kumanogoh, A., Ta-naka, T., Fujiwara, H., Hirata, M., Yamagami, T., Nakahata, T., et al.(1996). Targeted disruption of gp130, a common signal transducerfor the interleukin 6 family of cytokines, leads to myocardial andhematological disorders. Proc. Natl. Acad. Sci. USA 93, 407–411.

GenBank Accession Number

The nucleotide sequence data reported in this paper have beendeposited in the DDBJ/EMBL/GenBank nucleotide sequence data-base with the accession number AB028048.