Developmental Cell 11, 519–533, October, 2006 ª2006 Elsevier Inc. DOI 10.1016/j.devcel.2006.08.001
Clonal Analysis of Mouse Development Reveals
a Polyclonal Origin for Yolk Sac Blood Islands
Hiroo Ueno1,* and Irving L. Weissman1,2
1Institute of Stem Cell Biology and Regenerative
Departments of Pathology and Developmental Biology
Stanford University School of Medicine
Stanford, California 94305
Direct clonal analysis of tissue and organ maturation
in vivo is a critical step in the interpretation of in vitro
cell precursor-progeny relationships. We have devel-
oped a method to analyze clonal progenitor contribu-
tions in vivo using ES cells stably expressing separate
fluorescent proteins and placed into normal blasto-
cysts to form tetrachimeras. Here we applied this
method to the analysis of embryonic yolk sac blood is-
lands. In most vertebrates, yolk sac blood islands are
dothelial cells. It has been proposed that these line-
ages arise from a common clonal progenitor, the he-
mangioblast, but this hypothesis has not been tested
directly in physiological development in vivo. Our
analysis shows that each island has contributions
from multiple progenitors. Moreover, contribution by
individual hemangioblast progenitors to both endo-
thelial and hematopoietic lineages within an island, if
it happens at all, is an infrequent event.
The initial attempts to describe dynamic events in mam-
malian in utero development, and avian in ovo develop-
ment, were primarily microscopic observations on fixed
tissues. Later, this was aided by the production of allo-
phenic mice, a method of making aggregation chimeras
that were implanted into the uterus (Mintz and Silvers,
1967), and by injection of blastocyst inner cell mass
(ICM) and postimplantation cells (Rossant et al., 1978);
both used enzymatic markers and resolution at the level
of every cell was not possible. Production of mice with
knockout of specific genes often leads to dramatic phe-
notypes; they are useful to study gene function, but can-
clonality of tissue development. Promoter-driven re-
combinase (Cre and Flp) (Kos, 2004; Garcia-Otin and
Guillou, 2006) lineage marking is a dramatic help, but
also does not define the steps of development in line-
ages or clonality of lineages. Here we develop a general
method to produce multiorigin mouse chimeras by pro-
ducing separate ES lines that harbor and express differ-
ent fluorescent protein-encoding genes (EGFP, ECFP,
and mRFP1) (Campbell et al., 2002) knocked into the
Rosa locus (Zambrowicz et al., 1997) and injecting the
three different colored ES cells into wild-type blasto-
cysts just prior to implantation. This method can be
modified by recombinase-mediated promoter marking
of multicolor targets. Here we apply this model to the
analysis of the origin of the first sites of blood and
blood vessel generation—the early yolk sac blood is-
lands—and extend the analysis to determine whether
such blood islands derive from local, perhaps clonal,
Both hematopoietic and endothelial cells originate
from the mesoderm. During mouse embryonic develop-
ment the first mesodermal cells emerge at the early
streak stage, which are reported to give rise to embry-
onic and extraembryonic mesoderm (Tam and Bedding-
ies are consistent with the model that the progenitors of
yolk sac blood islands exist within the distal proximal
portion of the primitive streak and that commitment to
these progenitors begins as early as the mid streak
stage (Lawson et al., 1991; Kinder et al., 1999; Palis
et al., 1999; Huber et al., 2004). The progenitors would
then migrate to the extraembryonic yolk sac. Flk1 plays
an indispensable role in the generation of yolk sac blood
island cells and also intraembryonic hematopoiesis and
vasculogenesis; Flk1 knockout mice develop neither
blood nor blood vessels and die as embryos (Shalaby
et al., 1995, 1997). The first blood island appears in the
yolk sac at the neural plate stage (Haar and Ackerman,
1971; Kaufman, 1995). Electron microscopic analyses
of the yolk sac at this stage revealed that the mesoder-
mal cell mass appears within the connective tissue be-
tween the visceral endoderm and the mesothelial cell
layer, where it proliferates. Elongated cells appear at
the periphery of the mesodermal cell mass, and they
show some of the characteristics of the endothelium of
vitelline vessels. On the other hand, ferritin clusters ap-
pear in the cytoplasm of some cells located centrally in
the mesodermal cell mass. These cell clusters then dif-
ferentiate into at least erythroblasts and are enclosed
by the endothelial cell layer (Sasaki and Kendall, 1985).
These erythroblasts are nucleated and have fetal-type
hemoglobin (Barker, 1968; Brotherton et al., 1979).
Within the blood islands, many mitotic cells are ob-
served (Sasaki and Kendall, 1985). These observations
suggest that each blood island is derived from a re-
stricted number of progenitors. The blood islands prolif-
erate, merge, and form yolk sac vascular networks and
the vitelline vessels that connect with embryonic ves-
sels; blood circulation establishes at approximately
E8.25 (McGrath et al., 2003).
We have shown that the progeny of yolk sac blood
islands injected orthotopically and synchronically con-
tribute to lifelong hematopoiesis and lymphopoiesis
(Weissman et al., 1977, 1978; summarized in Akashi
and Weissman, 2001). Separate origins of yolk sac and
intraembryonic hematopoiesis have also been postu-
lated (Cumano et al., 1996, 2001; Medvinsky and Dzier-
zak,1996),but notbytransplantationexperiments orlin-
eage marking invivo. Therefore, to understand thekinds
ofstemcells involved inthe generation ofyolk sacblood
islands, and the fate of their progeny in postembryonic
2Lab address: http://weissmanlab.stanford.edu
life, we believe that establishment of experimental
methods to follow clonogenic precursors in vivo such
as described above are required.
The term hemangioblast was introduced to describe
cells within chick embryo cultures that appear to pro-
duce both hematopoietic and endothelial lineages (Mur-
ray, 1932). This concept was supported by histological
observations that hematopoietic and endothelial cells
of each blood island originate from the same population
of cells, that is, the mesodermal cell mass, and that they
develop in close spatial and temporal proximity (Haar
and Ackerman, 1971). Moreover, studies on in vitro dif-
ferentiationof mouse embryonic stem(ES) cells have re-
vealed that during differentiation from ES cells to hema-
topoietic cells, clonal progenitors that can generate
both endothelial and hematopoietic cells are generated
mangioblasts express the cell surface receptor Flk1,
and flk1 double mutant mice are embryonic lethals
with no appearance of blood or blood vessels (Shalaby
et al., 1997). On the basis of these observations, it was
reasonable to propose that hemangioblasts are the ex-
clusive progenitors that can generate both cell lineages
(Robertson et al., 1999). Supporting this, it has been
shown that within the brachyury- and Flk1-positive frac-
there are cells that differentiate into both hematopoietic
and endothelial cells, at least in vitro, and it was pro-
posed that these were the hemangioblasts (Huber
et al., 2004). In these experiments a single cell gave
rise to cells of both fates, demonstrating the bipotency
of the progenitor.
In order to develop a general method to assess clono-
genic cells in the developing mouse embryo, we estab-
marked by three different fluorogenic proteins ex-
pressed through knockin to the Rosa locus were used
in blastocyst injection. These could be injected as single
cells or multiples of each fluor type (EGFP, ECFP, and
mRFP1), and their progeny followed along with host un-
colored ICM cells. In some cases, lineal descendants
can be marked by a Promotor-cre gene coexisting in
the ES cell with a sensitive floxed color or multicolor re-
porter construct. Using these methods, we show here
that each yolk sac blood island is derived from multiple
clonal precursors, that for the most part these precur-
sors are not bipotent hemangioblasts contributing to
the same island, and that most blood cells do not derive
from an Flk1-positive precursor.
Establishment of Multicolor, Multiclonal Mice Using
Rosa26 Knockin Fluorescent ES Cells
We established mouse ES clones expressing three
different fluorescent genes, EGFP, ECFP, and mRFP1
(Campbell et al., 2002). These marker cDNAs were intro-
duced into the Rosa26 locus by a knockin strategy
(Zambrowicz et al., 1997) (see Figures S1A and S1B in
the Supplemental Data available with this article online).
Marker genes integrated into the Rosa26 locus are sta-
bly expressed in virtually all tissues (Zambrowicz et al.,
1997). Although in this strategy only a single copy of
the gene can be introduced into the cell at a time, strong
expression of the marker genes can be achieved by in-
serting the chicken b actin promoter with cytomegalovi-
et al., 1989) with the cDNA (Zong et al., 2005). This
method is advantageous because blood islands in the
yolk sac have a relatively high degree of autofluores-
cence (data not shown). The resulting clones, Rosa-
EGFP, Rosa-ECFP, and Rosa-mRFP1 cells, expressed
high levels of fluorescent proteins and the expression
of these fluorescent proteins did not affect the prolifera-
tion and maintenance of multipotency of the ES cells
in vitro (Figure 1A and data not shown). Because their
emission wavelengths overlap with each other, we con-
firmed that the filter set of immunofluorescence micros-
copy used in this study could distinguish between their
fluorescence (Figure 1A). To confirm that fluorescent
markers are not downregulated in differentiated cell
types in this knockin strategy, a mixture of three differ-
ent fluorescent-marked ES clones was injected into
E3.5 blastocysts (Figure 1B), and chimeric embryos
and adult mice of Rosa26 EGFP, ECFP, or mRFP1
knockin ES clones were generated. Genomic PCR of
FACS-separated peripheral blood cells (Figures S2A–
S2C) and total bone marrow cells (Figures S2D and
S2E) from the chimeric mice revealed that cells carrying
the knockin allele were detected only in the fluorescent-
positive fraction and not in the fluorescent-negative
fraction of the cells (Figures S2C and S2E).
Early Mouse Embryos Show the Contribution of Four
Different Pluripotent Cells to All Embryonic
and Developing Embryo-Derived Extraembryonic
In these experiments, fluorescent-marked cells distrib-
uted in all the tissues in chimeric embryos (E7.5–E18.5)
(Figures 1C–1E and data not shown) and adult chimeric
mice we examined (Figures 1F–1H and data not shown).
These results indicate that the marker genes are stably
expressed in differentiated cells in vivo and the possibil-
ity of downregulation of fluorescent genes was unlikely,
especially in hematopoietic lineages.
In this experiment, the resulting chimeric embryos
originated not only from the injected ES cells but also
from the host-derived ICM. Thus, the embryos would
be composed of four types of cells: three types of fluo-
rescent cells and one type of nonfluorescent cell. Before
the initiation of the blood circulation, the migration of
cells should be localized. Therefore, if the cells that are
adjacent to each other are of the same origin, they
having different origins can either be of different colors
or show the same color. However, by repeating the ex-
periments and examining multiple blood islands, it
would be almost impossible that cells would always
show the same color if they are adjacent to each other
but have a different origin. As an example of this princi-
ple, in analyzing adult chimeric mice, we observed that
cell clusters within the epithelium of the small intestine,
actually complete villi, show only a single color,whereas
the lamina propria with blood cells and the mesenchy-
mal and smooth muscle layers were composed of three
sult, we could confirm that the tissues of clonal origin
show single-color cell clusters in our system. However,
if an embryo had unbalanced chimerism and the tissues
were biased toward one color, it often happened that
cells adjacent to each other but with different origins
displayed the same fluorescence (data not shown).
Based on this observation, we decided to analyze sam-
colors in this assay. In summary, these results indicate
that the chimeric analysis used in this study is useful
to examine whether tissues of interest are monoclonal
or polyclonal and whether cell clusters that are adjacent
to each other have the same origin or not.
Analysis of Yolk Sac Blood Islands
A mixture of three different fluorescent-marked ES
clones was injected into E3.5 blastocysts (Figure 1B),
and chimeric embryos were obtained and analyzed
4.5 days after injection. We observed that the develop-
ment of these embryos was delayed to various degrees
(from 12 hr to 48 hr) (data not shown). This might be due
to processes inherent in ES cell chimeras, to the effect
of the blastocyst injection, and/or the overexpression
of fluorescent proteins in the ES cells. As a result, the
developmental stages of injected embryos varied
Figure 1. Establishment of Clonal Analysis of Embryos and Adult Mice Using Fluorescent ES Cells
(A) Fluorescent marker expression of ES clones. A mixture of Rosa-EGFP, -ECFP, and -mRFP1 clones was cultured on mouse embryonic feeder
cells and observed with confocal microscopy.
(B) Blastocysts injected with Rosa-EGFP, -ECFP, and -mRFP1 clones. Blastocysts were injected with a mixture of Rosa-EGFP, -ECFP, and
-mRFP1 ES cells and were incubated in M16 medium for 3 hr, fixed in 4% paraformaldehyde, and observed. Arrows indicate holes opened
by an injection needle.
(C) A whole-mount image of a late headfold (LHF) stage chimeric embryo obtained in this study. This embryo is an ECFP, mRFP1, and nonflu-
orescent cell chimera.
(D) Caudal part of midbrain of an E14.5 chimeric embryo.
(E) Liver of an E14.5 chimeric embryo. Hematopoietic cells with three colors are observed.
(F) Section of adult cerebellar cortex (3 weeks), as an example of ectodermal tissues. Open triangles indicate Purkinje neurons.
(G) A section of adult bone marrow (3 weeks), as an example of mesodermal tissues.
(H) A section of adult small intestine (5 weeks), as an example of endodermal tissues. Left panel: small intestine epithelial cells derived from
a clonal progenitor form cell clusters with a single color. Note that lamina propria and smooth muscle layers are composed of a mixture of three
colors. e, epithelium; lp, lamina propria; sm, smooth muscle cell layer. Right panel: Hoechst 33342 staining of nucleus of the same section, in-
dicating that single-color epithelium is composed of multiple cells.
(A–H) Red, mRFP1-expressing cells; green, EGFP-expressing cells; blue, ECFP-expressing cells. Merged images are shown.
The scale bars represent 25 mm (A and B); 200 mm (C); 100 mm (D–F and H); and 50 mm (G).
Origin of Yolk Sac Blood Islands
more than those of wild-type embryos at the same
stage. However, because this never caused faster em-
bryonic development, all of the embryos were at the
stages before the somite stage (data not shown), well
before the establishment of the connected embryonic
and extraembryonic vasculature. Judging from their
morphology, the embryos obtained were equivalent to
those from the mid streak (MS) to the late headfold
(LHF) stage (the staging defined by Downs and Davies,
1993). Figure 2A illustrates the schematic of expected
results of yolk sac blood islands analyzed by this strat-
egy. The progenitors of the yolk sac blood islands ap-
pear within the distal proximal region of the primitive
streak (Huber et al., 2004; Kinder et al., 1999; Lawson
et al., 1991), which migrate to the yolk sac. Morphoge-
netic analysis of early embryos has indicated that each
blood island is generated from a limited number of pro-
genitors (Sasaki and Kendall, 1985). We classified
blood islands into type I to type IV as shown in Fig-
ure 2A. Considering the possibility that they are derived
from both bipotential and lineage-committed progeni-
tors, we subclassified type IV blood islands into type
IVa to type IVc (Figures 2B–2D). Apart from types I–IV
there are several other possibilities, such as blood is-
lands derived from monoclonal endothelial and poly-
clonal hematopoietic progenitors, or those originated
from a mixture of bipotential progenitors and monopo-
Figure 2. Classification of Blood Islands
(A)Schematicof expected results of yolksacblood islandsobtained in thisstudy. Light blue arrows denote migration from the primitive streak to
the yolk sac. In the case wherein putative bipotential progenitors migrate to the yolk sac and each blood island is generated from a single bipo-
tential progenitor, endothelial cells and hematopoietic cells within a blood island should show a single color (type I). In the case wherein putative
bipotential progenitors proliferate in the primitive streak and mix during migration to the yolk sac, both endothelial and hematopoietic cells
should be chimeric, but both would always show the same mixture of colors (type II). In the case wherein the endothelial and hematopoietic pro-
genitorsareseparatelygeneratedintheprimitivestreakandeachblood islandis derivedfromasingle hematopoieticandendothelial progenitor,
bothhematopoietic clusters andendothelial celllayerswould show a single color, butthe lineages could either beadifferent color(type III)orthe
same color (type I). In the case wherein the endothelial and hematopoietic progenitors are separately generated in the primitive streak, and
each blood island is derived from several hematopoietic and endothelial progenitors, the blood islands would be similar to the image indicated
in type IV.
(B–D) Subclassification of type IV blood islands.
defined as type IVa.
(C) If those in the endothelial cells are a subgroup of those in the hematopoietic cells, they are subclassified as type IVb.
(D) If some colors do not overlap in both lineages, they are classified as type IVc.
The results of the experiment are shown in Figure 3
and Table S1. We confirmed that all three of the ES
clones could contribute tothe formation of chimeric em-
bryos at this stage (Figure 3A). We analyzed embryos
contributions from two or three different colored precur-
sors. The maturity of examined blood islands varied
within an embryoand oneembryocouldhave
Figure 3. Chimeric Analysis of Yolk Sac Blood Islands with Fluorescent ES Clones
(A) A chimeric embryo at the early neural plate stage (EB). A white rectangle indicates an immature blood island (an arrow), whose magnified
picture is shown in (B). The arrow indicates the direction of tissue extension. ec, embryonic ectoderm; me, intraembryonic mesoderm; ve, vis-
ceral endoderm; am, amnion; al, allantois; ch, chorion.
(B and C) An immature blood island with EGFP-, ECFP-, or mRFP1-expressing and nonfluorescent progenitor cells.
(B) Magnified picture of (A), white rectangle.
(D–F) Type II blood islands whose endothelial and hematopoietic cells express EGFP, ECFP, and mRFP1.
(G and H) Type IVa blood islands.
(G) A type IVa blood island with EGFP, ECFP, and mRFP1 chimeric endothelial cells and ECFP and mRFP1 chimeric hematopoietic cells.
(H) A type IVa blood island with EGFP, ECFP, and mRFP1 chimeric endothelial cells and ECFP, EGFP, and nonfluorescent chimeric hematopoi-
(I and J) Type IVb blood islands.
(I) A type IVb blood island with EGFP and mRFP1 chimeric endothelial cells and EGFP, ECFP, and mRFP1 chimeric hematopoietic cells.
(J) A type IVb blood island with ECFP and mRFP1 chimeric endothelial cells and EGFP, ECFP, mRFP1, and nonfluorescent chimeric hematopoi-
(K–M) Type IVc blood islands.
(K) A type IVc blood island with EGFP, ECFP, and nonfluorescent chimeric endothelial cells and ECFP, mRFP1, and nonfluorescent chimeric
(L) A type IVc blood island with EGFP, ECFP, and nonfluorescent chimeric endothelial cells and ECFP, mRFP1, and nonfluorescent chimeric
(M) A type IVc blood island with ECFP, mRFP1, and nonfluorescent chimeric endothelial cells and EGFP, ECFP, and nonfluorescent chimeric
(N) Distribution of fluorescent-expressing cells to the endothelial and hematopoietic lineages. NS, not significant as assessed by ANOVA.
(O) EGFP-, ECFP-, and mRFP1-expressing cells are equally contributed to the endothelial and hematopoietic cells in the blood islands. NS, not
significant as assessed by ANOVA.
(A–L) Red, mRFP1-expressing cells; green, EGFP-expressing cells; blue, ECFP-expressing cells. Merged images are shown. VE, visceral endo-
derm. The scale bars represent 25 mm.
Origin of Yolk Sac Blood Islands
fluorescent-positive and fluorescent-negative blood is-
lands at the same time (data not shown). In immature
blood islands, blood cells and endothelial cells could
not be distinguished by morphology (Figures 3B and
3C). At this stage of embryos, it has been reported that
lineage commitment of progenitors could be assessed
by in vitro colony assay (Palis et al., 1999; Drake and
Fleming, 2000). However, the method cannot be applied
were not included in statistical analyses. In this study,
however, all of the immature blood islands found in
two- or three-color embryos had two or three different
fluorescent-marked contributors, indicating that they
are derived from polyclonal progenitors (Figures 3B
and 3C and data not shown). In these experiments, the
number of blood islands analyzed was one to nine per
embryo. We obtained a total of 80 embryos. We ex-
fluorescence contribution, and embryos with somites.
Among these, 38 were fluorescent embryos with blood
islands, 24 of which had contributions from two or three
different fluorescent cells. We sectioned all embryos
from end to end and confirmed that no somite formation
had occurred in the samples we analyzed, indicating
that blood circulation had not formed in these embryos,
and that the blood islands we analyzed were not con-
obtained, 31.0% were type II blood islands (Figures 3D–
3F) (polyclonal with blood and endothelial cells of the
same combination of color) and 69.0% were type IV
blood islands (polyclonal with blood and endothelial
cells of a different combination of color) (Figures 3G–
3M; Table 1; Table S1). Within type IV blood islands,
type IVa (Figures 3G and 3H) blood islands were most
frequent, although there were also considerable num-
bers of type IVb (Figures 3I and 3J) and IVc (Figures
3K–3M) blood islands (Table 1). Type I blood islands
were absent, indicating that no blood islands derived
from aclonal migrating hemangioblast. TypeIII blood is-
lands were also absent, thereby indicating that each
blood island is derived from several independent (poly-
clonal) progenitors. Blood islands other than type I to
type IV were not detected, thereby indicating that the
other minor possibilities raised above could be ruled
out. Fluorescent cells were equally distributed in both
endothelial and hematopoietic lineages (Figure 3N) and
three fluorescent clones equally contributed to blood is-
lands (Figure 3O; Table 1). In conclusion, these results
indicate that the direct progenitors of each blood island
are polyclonal and generally have a single fate. In fact, in
Figure 3A, the arrow points to an early developing struc-
ture with the location and morphology of an incipient
blood island, and it is clear that several independent
precursors are present.
Single-Cell Injection of Rosa Knockin ES Clones
In these experiments the injected ES cells are incorpo-
rated into the ICM, from which the chimeric embryos
are generated, and, therefore, the chimeric patterns
should be similar in all portions of the embryos. Thus,
it is possible that the type II blood islands, indicated in
Table S1, include blood islands in which endothelial
and hematopoietic cells are derived from different cell
clusters. To test that possibility, we next performed
single-cell injection of each of the Rosa26 knockin
Table 1. Summary of Yolk Sac Blood Islands by Clonal Analysis
(One Cell Each)
Total embryos analyzed
Blood islands with fluorescence
Fluorescence contribution to blood cells
Fluorescence contribution to endothelial
Injected ES clones contribution to blood
InjectedES clonescontribution toendothelial
Embryos with hematopoietic specific
Embryos with endothelial specific
NA NANA 38c
4 (16.7%) 0 (0.0%)NANA
7 (29.2%)7 (46.7%)NANA
NA, not applicable; NS, not significant by ANOVA.
aEmbryos with two or three fluorescences were analyzed.
bp < 0.01 by ANOVA.
cNumbers of blood islands to which fluorescent cells or b-galactosidase-positive cells contribute. Blood islands in which b-galactosidase
immunostaining was not performed were excluded.
fluorescent clones. In this experiment, we expected
that, by decreasing the degree of chimerism, asymmet-
ric distribution of single ES cell-derived cells within an
embryo would be more apparent.
In this experiment, chimeric embryos obtained had
one or two fluorescent cells but none had three colors
(Table S2). The levels of chimerism were considerably
decreased and cells distributed in some specific line-
ages were more frequently observed than in the experi-
ments of Figure 3 (data not shown). We included em-
bryos with one fluorescence in this analysis and the
results showed that type IV blood islands made up
86.4% of the blood islands, and the rest (13.6%) were
type II (Figure 4; Table 1; Table S2). Of the 57 type IV
blood islands, there were 40 type IVa blood islands
(60.6% of total) (Figures 4A and 4B; Table 1), much
more frequent than type IVb (Figures 4C–4E) and type
IVc (Figures 4F and 4G) blood islands. Of the 40 type
IVa blood islands, 31 contained fluorescent endothelial
and nonfluorescent blood cells, and this type of blood
island tended to be repeatedly observed within the
same embryo. We also noticed that in such embryos,
hematopoietic cells of the same color as endothelial
cells were not found. It appears that these endothelial
cells originated from endothelial specific progenitors
and were not derived from local hemangioblasts (by
our definition). To examine this, the fluorescence contri-
bution to blood islands was summed with each embryo
(whole-embryo analysis) (Table S3). In this analysis, we
found that within 46.7% of embryos, one fluorescence
only contributed to the endothelial lineage (Tables S1–
S3, blue letters; Table 1; Figures 4H and 4I). Retrospec-
tively, there were also many instances (29.2%) that one
fluorescence contributed specifically to the endothelial
lineage in the experiments of Figure 3 (Table 1; Table
S1). If all of the endothelial cells within yolk sac blood is-
lands at this stage originate from hemangioblasts, there
must be hematopoietic cells of the same color. There-
fore,theseexamples provide evidencethatthereareen-
dothelial specific progenitors that are not derived from
the hemangioblast, unless in all cases the hemangio-
blasts had not reached the yolk sac membranes, and
in those cases all detectable hematopoietic cells disap-
peared. In these experiments, fluorescent cells prefer-
entially distributed in the endothelial lineage (Figure 4J),
whereas three fluorescent clones equally contributed
to blood islands (Figure 4K). These results strongly
suggest that the endothelial progenitors that are not de-
rived from the hemangioblast supply many endothelial
cells with blood islands. The difference was not ob-
served in Figure 3M, suggesting that high degrees of
chimerism caused by injection of multiple cells masked
Examples of embryos in which one color only contrib-
uted to blood cells were fewer than those of endothelial
cells (16.6% in Table S1 and none in Table S2). However,
there were some cases in which one color contribution
was considerably biased to the hematopoietic lineage
(Figures 4L and 4M and data not shown). In some cases,
blood islands with fluorescent endothelial and nonfluo-
rescentbloodcells andthose withnonfluorescentendo-
thelial and fluorescent blood cells coexist within an
embryo (Figure 4N). These results do not exclude the
possibility that most hematopoietic cells originate from
bipotential progenitors; these could be pre-dpc 7 meso-
dermal precursors at a distance from yolk sacs to ex-
plain the yolk sac results, and independent of yolk
sacs if intraembryonic blood cells are considered. How-
ever, the variety of distribution of fluorescent cells in
both lineages shown in this experiment suggests that
the origin of hematopoietic cells is heterogeneous.
Labeling of Blood Island Cells that Had Flk1-Positive
Mice that lack flk1 genes are embryonic lethals that lack
blood and blood vessels (Shalaby et al., 1997). Flk1 is
not only a proposed marker for the hemangioblast
(Huber et al., 2004) but also has been proposed as
a marker for both hematopoietic (Eichmann et al.,
1997) and endothelial (Kabrun et al., 1997; Drake and
Fleming, 2000) progenitors in yolk sac hematopoiesis,
although we could not show hematopoietic potential
of transferred mouse marrow Flk1-positive cells (Chris-
tensen, 2003). However, it is not clear whether all blood
and endothelial cells in blood islands that arise in devel-
opment originate from Flk1-positive progenitors. We
therefore wished to test this possibility directly. To do
so, we labeled Flk1-positive cells and their progeny as
they differentiate into hematopoietic and endothelial
progenitors in vivo by using cre/loxp lineage marking.
The first purpose of the experiment was to check
whether all the endothelial and hematopoietic cells orig-
inate from Flk1-positive progenitors. The second pur-
pose of the experiment was to limit our examination to
cells that had expressed Flk1, ignoring cell mixing that
might have derived from Flk1-negative progenitors.
We constructed fluorescent expression vectors (loxp-
stop-loxp-EGFP and loxp-stop-loxp-ECFP) as shown
in Figure 5A. We separately introduced the vectors into
Flk1-cre knockin ES cells (Motoike et al., 2003) and es-
tablished clones. The clones Flk1-cre/loxp-stop-loxp-
EGFP and Flk1-cre/loxp-stop-loxp-ECFP were mixed
and injected into the blastocysts, and the blood islands
were analyzed. In this system, the progeny of the ES
cells should be marked with fluorescent proteins after
they have differentiated into Flk1-positive progenitors
in vivo. The results are shown in Figures 5B–5E, Table 1,
andTable S4. We found that 74.5% ofthe yolk sacblood
islands were type IV and the rest were type II (Table 1;
Table S4). It is noteworthy that fluorescent-labeled cells
were preferentially distributed in the endothelial lineage
(Figure 5F; Table S4), whereas EGFP-expressing cells
and ECFP-expressing cells were equally distributed in
the blood islands (Figure 5G; Table S4). Because the
fluorescent expression vectors were not introduced
into the Rosa locus in this experiment, it was possible
that partial downregulation of fluorescent genes in he-
matopoietic cells occurred. To rule out this possibility,
we put another marker gene, b-galactosidase, between
two loxp sites of the fluorescent vectors as shown in
Figure 5H. In this strategy, injected ES cell-derived cells
express b-galactosidase unless recombined with cre.
The constructs were confirmed by transient expression
in Cos7 cells (Figure S3). We separately introduced
these vectors into Flk1-cre ES cells, established clones,
injected these clones into blastocysts, and analyzed the
resulting blood islands. In this experiment, fluorescent-
labeled cells were also preferentially distributed in the
Origin of Yolk Sac Blood Islands
endothelial lineage (Figure 5I; Table S5), whereas EGFP-
expressing cells and ECFP-expressing cells were
equally distributed in the blood islands (Figure 5J; Table
S5). By immunostaining of b-galactosidase (Figures 5K–
5M), we could confirm that injected ES cell-derived cells
equally distributed in both lineages (Figure 5N; Table 1;
Table S5), suggesting that partial downregulation of
marker genes in blood cells is unlikely in these experi-
ments. In this experiment, because numbers of ES cells
injected into blastocysts were the same as those in
Figure 4. Chimeric Analysis of Yolk Sac Blood Islands by Single-Cell Injection of Each of the Fluorescent ES Clones
(A) A type IVa blood island with EGFP and nonfluorescent chimeric endothelial and nonfluorescent hematopoietic blood cells.
(B) A type IVa blood island with EGFP, ECFP, and nonfluorescent chimeric endothelial and nonfluorescent hematopoietic blood cells.
(C) A type IVb blood island with nonfluorescent endothelial and ECFP and nonfluorescent chimeric hematopoietic cells.
(D) A type IVb blood island with nonfluorescent endothelial and EGFP and nonfluorescent chimeric hematopoietic cells.
(E) A type IVb blood island with nonfluorescent endothelial and ECFP and EGFP and nonfluorescent chimeric hematopoietic cells.
(F) A type IVc blood island with EGFP and nonfluorescent chimeric endothelial and mRFP1 and nonfluorescent chimeric hematopoietic cells.
(G) A type IVc blood island with ECFP and nonfluorescent chimeric endothelial and EGFP and nonfluorescent chimeric hematopoietic cells.
(H) Two type IVa blood islands showing the same chimeric pattern, with EGFP and nonfluorescent chimeric endothelial and nonfluorescent
hematopoietic blood cells.
(I) Two type IVa blood islands showing the same chimeric pattern, with mRFP1 and nonfluorescent chimeric endothelial and nonfluorescent
hematopoietic blood cells.
(J) Fluorescent-expressing cells are preferentially distributed in the endothelial lineage with statistical significance as assessed by ANOVA.
(K) EGFP-, ECFP-, and mRFP1-expressing cells are equally distributed to the endothelial and hematopoietic cells in the blood islands. NS, not
significant as assessed by ANOVA.
(L and M) Examples of blood islands in embryos in which fluorescent markers preferentially distribute in blood cells.
(L) Type II blood islands with mRFP1 and nonfluorescent chimeric endothelial and hematopoietic cells, in which fluorescence contribution is
biased to the hematopoietic lineage.
(M) Type II blood islands with EGFP, mRFP1, and nonfluorescent chimeric endothelial and hematopoietic cells, in which fluorescence contribu-
tion is biased to the hematopoietic lineage.
(N) Three mRFP1 and nonfluorescent chimeric blood islands. One is a type IVb blood island in which mRFP1 contributes only to hematopoietic
cells. Others are type IVa blood islands in which mRFP1 contributes only to endothelial cells.
(A–IandL–N)Red, mRFP1-expressing cells;green,EGFP-expressingcells; blue,ECFP-expressing cells. Merged imagesare shown. VE, visceral
The scale bars represent 25 mm (A–G, L, and M); 50 mm (H and I).
Figure 3, the result of Figure 5N corresponds to that of
Figure 3N. Thus, the results of Figures 5F and 5I do not
correspond to that in Figure 4J. These results indicate
poietic cells than in endothelial cells because many of
the hematopoietic cells were not recombined with cre.
There are two situations that can explain these results.
First, some of the hematopoietic cells are derived from
Flk1-low or -negative progenitors. Second, progenitors
of hematopoietic cells only transiently express Flk1,
while endothelial progenitors keep expressing Flk1.
To distinguish between these possibilities, we per-
formed follow-on experiments as shown in Figure 6.
We constructed a fluorescent expression vector (loxp-
CDG) as shown in Figure 6A. On treatment with cre,
which put the DsRED2, ECFP, or EGFP cDNA in a posi-
tion immediately downstream of the actin promoter. The
construct was confirmed by transient expression into
Cos7 cells (Figure 6B). The vector was introduced into
Flk1-cre ES cells, and the resulting clone was injected
into blastocysts; the blood islands were analyzed. In
rived cells was low, the cells would remain nonfluores-
cent or express ECFP, DsRED2, or EGFP due to partial
loxp recombinations. If Flk1 expression in the cells
was high enough, most cells would turn EGFP positive,
because all of the loxp sites would be recombined
(Figure 6C). In this system, if there is low expression of
Flk1-cre, the result is expected as shown in Figure 6C,
should be as presented in Figure 6C, model 2. The re-
sults ofthe experiments showed that onlyEGFP-labeled
cells were found in hematopoietic and endothelial cells
in all of the blood islands we analyzed (Figures 6D and
6E, and data not shown). These results, together with
the results in Figure 5, which show that many of the he-
the model that blood cells in the yolk sac blood islands
originate fromseparateFlk1-high andFlk1-low/negative
progenitors. These results can be taken to suggest that
the origins of hematopoietic cells were heterogeneous.
In these experiments, all endothelial cells are derived
from Flk1-high progenitors, whereas most blood cells
are not. If expression of Flk1 is a constant marker of he-
mangioblasts, the clone-derived blood cells that were
not colored did not come from hemangioblasts. We
also confirmed from the results of classification of the
blood islands in these experiments that the progenitors
of the blood islands were polyclonal and generally have
a single fate.
Chimeric Analysis by Blastocyst Injection
Lineage-tracing experiments, by orthotopic transplan-
tation of early embryonic cells having different markers,
have been performed since the 1970s. Chimeric analy-
ses by blastocyst injection have been previously per-
formed (Gardner and Rossant, 1979). This method is
useful to analyze the origins of hematopoietic and endo-
thelial progenitors for yolk sac hematopoiesis, because
of the model that clonal and bipotential progenitors
generate both lineages in the yolk sac blood islands.
Markers used for the analysis of subsequent chimeras
were originally the expression of allelic forms of gluco-
phosphatase isomerase (GPI) (Gardner and Rossant,
1979), tracking with locally infused lipophilic dyes such
as Dil and DiO (Serbedzija et al., 1989), and expression
of horseradish peroxidase (Lawson et al., 1991). How-
ever, they cannot be directly assayed in situ, and there-
fore these analyses have not been convenient for han-
dling many samples or do not give histological detail.
By utilizing EGFP, its variant ECFP, and the recently de-
veloped less toxicmutant of RFP (Campbell etal., 2002),
we succeeded in developing a general method for histo-
logical identification of green, red, and blue cells in all
tissues, in which one could statistically analyze a large
number of chimeric embryos, their various tissues and
organs, and as the major target of this report, yolk sac
blood islands. Moreover, we showed in this study that
the strategy could be combined with the cre-loxp cell-
marking system, which enables us to label cells at the
progenitor level as long as appropriate promoters or
cre-knockin cells specific to the specific lineages of in-
terest are available. The strategy is relevant for studying
the developmental origins of any organ or tissue in mice
and should have wide applicability.
What Can Be, and Cannot Be, Concluded
from the Single-Cell Injection Experiments
The ICM cells within E3.5 blastocysts are composed of
approximately 20 cells and each cell theoretically has
pluripotency and can contribute to any kind of tissue
cells. It has also been reported that epiblasts at E4.5
possess pluripotency at the single-cell level (Gardner
and Rossant, 1979; Gardner et al., 1985); however, the
ability is lost after E5.5 (Rossant et al., 1978). Thus, in-
jected ES cells should be able to expand while retaining
pluripotency for at least 24 hr. The doubling time of epi-
blasts at this stage is as short as 4.4 hr (Snow, 1977).
From these findings it is important to note that, in sin-
gle-cell injection experiments, it is always possible that
cells of the same fluorescence are generated from cells
of distinct lineages. However, it is safe to say that cells
that have different colors are derived from different pro-
genitors. By decreasing the number of cells injected, we
could obtain embryos with lower levels of chimerism
(Table S2). Fortunately, this uncovered the fact that sin-
gle ES cell-derived cells tended to distribute asymmetri-
cally in the embryo, which had been masked by higher
chimerism caused by multiple-cell injection. In the
whole-embryo analysis, we found many examples in
which one fluorescence only contributed to the endo-
thelial lineage. If injected ES cells are already committed
to the endothelial lineage, they cannot contribute to the
embryo (Rossant et al., 1978), because the endothelial
committed cells are equivalent to the progenitors found
in E7.0 embryos (Drake and Fleming, 2000). Therefore,
it clearly shows that these cells are not derived from
Direct Progenitors of Yolk Sac Blood Islands
As described above, the original observation that sup-
ported the hypothesis that endothelial and hematopoi-
etic cells share a common precursor was the spatially
and temporally close developmental association of
these two lineages in mammals and avian species
Origin of Yolk Sac Blood Islands
Figure 5. Chimeric Analysis of Yolk Sac Blood Islands Using the Flk1-cre Labeling System
(A) Schematic of loxp-stop-loxp-EGFP and loxp-stop-loxp-ECFP vectors.
(B and C) A type II blood island whose hematopoietic and endothelial cells are composed of EGFP, ECFP, and nonfluorescent chimera.
(D) A type IVa blood island with ECFP and nonfluorescent chimeric endothelial cells and nonfluorescent hematopoietic cells.
(E) A type IVa blood island with EGFP, ECFP, and nonfluorescent chimeric endothelial cells and EGFP and nonfluorescent chimeric hematopoi-
(B–E) Green, EGFP-positive cells; blue, ECFP-positive cells; red, nonspecific signals emitted to the red channel (see Experimental Procedures).
Merged images are shown.
(F) Fluorescent-expressing cells are preferentially distributed in the endothelial lineage with statistical significance as assessed by ANOVA.
assessed by ANOVA.
(H) Schematic of loxp-lacZ-loxp-EGFP and loxp-lacZ-loxp-ECFP vectors.
(I) Fluorescent-expressing cells are preferentially distributed in the endothelial lineage with statistical significance as assessed by ANOVA.
(J)EGFP- andECFP-expressing cellsare equallydistributed intheendothelial andhematopoietic cellsintheblood islands. NS,not significantas
assessed by ANOVA.
(K–M) Anti-b-galactosidase staining of the blood islands. Green, EGFP-positive cells; blue, ECFP-positive cells; red, b-galactosidase-positive
cells. Merged images are shown.
(K) A type IVa blood island with ECFP and nonfluorescent chimeric endothelial cells and nonfluorescent hematopoietic cells with b-galactosi-
dase-positive hematopoietic cells.
(L) A type IVa blood island with EGFP and nonfluorescent chimeric endothelial cells and nonfluorescent hematopoietic cells with b-galactosi-
dase-positive endothelial and hematopoietic cells.
(M) A type IVa blood island with EGFP and nonfluorescent chimeric endothelial cells and nonfluorescent hematopoietic cells. Some of the
endothelial cells were b-galactosidase positive.
(Haar and Ackerman, 1971; Robertson et al., 1999). Be-
fore establishment of blood circulation, blood cell clus-
ters are enclosed by endothelial layers and the growing
blood islands become separated from each other. This
finding led to the hypothesis that both lineages in each
blood islands are generated from local single or multiple
clonal hemangioblasts has been excluded. Even if we
suppose that they could originate from a combination
of hemangioblasts and endothelial specific progenitors,
the generation of type IVb and IVc blood islands (Tables
S1 and S2) could not be explained. Moreover, 31 out of
40 type IVa blood islands in Table S2 were combinations
(N) Injected ES cell-derived cells, defined by expression of fluorescent proteins or b-galactosidase, are equally distributed in the endothelial and
hematopoietic cells in the blood islands. NS, not significant as assessed by ANOVA.
(B–E and K–M) VE, visceral endoderm. The scale bars represent 25 mm.
Figure 6. Labeling of Yolk Sac Blood Islands Using Flk1-cre/loxp-CDG ES Cells
(A) Schematic of the loxp-CDG vector.
(B) Transientexpression of loxp-CDGandMx1-cre vectors in Cos7 cellsthat weretreated withPIPC. Inthiscase, more than one vector had been
introducedintocellsandcre-recombinedloxpsitesatrandom,resultinginexpression ofrandomcombinationoffluorescentproteins. Leftupper
panel: EGFP; right upper panel: ECFP; left lower panel: DsRED2; right lower panel: merged image.
(C) Schematic of cre-mediated fluorescent gene expression of the loxp-CDG vector. Upper panel: four loxp sites are recombined with cre at ran-
a least two distinct populations exist within blood cells, Flk1-high and Flk1-low/negative. Model 2: blood cells are composed of a single popu-
lation, with various degrees of Flk1 expression levels.
(D and E) Type II blood islands whose endothelial and hematopoietic cells are composed of EGFP and nonfluorescent chimera.
(B, D, and E) Red, DsRED2-positive cells; green, EGFP-positive cells; blue, ECFP-positive cells. Merged images are shown. VE, visceral endo-
derm. The scale bars represent 25 mm.
(F) Schematic of yolk sac blood island development. Model 1: clonal hemangioblast model. In this model, one hemangioblast always differen-
tiates intoone hematopoietic and one endothelial progenitor.Allthe hematopoietic andendothelialcellsin theyolk sac blood islands are derived
fromthe hemangioblast. Model2:the hemangioblastandendothelial specificprogenitor model.Endothelial andhematopoietic cellsinthe blood
islands are generated from bipotential progenitors, the hemangioblasts, and endothelial specific progenitors. The hematopoietic specific pro-
genitors are also suggested to exist. Model 3: bipotential hemangioplast model. In this model, the hemangioblasts can differentiate into endo-
thelial, hematopoietic, or both lineages. Light blue cells denote cells destined to endothelial lineage. Pink cells denote cells destined to hema-
topoietic lineage. White cells denote cells not committed to hematopoietic or endothelial lineages.
Origin of Yolk Sac Blood Islands
of fluorescent endothelial and nonfluorescent blood
cells (Table S2), which do not directly support that
they originate from hemangioblasts.
The exclusion of the clonal hemangioblast as the
in situ progenitor for each yolk sac blood island opens
the question of how many independent hematopoietic
and endothelial stem/progenitor cells colocalize to
form these blood islands in the absence of an estab-
lished vasculature. We propose that rare yolk sac cells
prior to the origin of blood islands produce and secrete
a chemokine(s) that establishes the localization of many
independent progenitors of the two lineages.
We propose that the close association of both line-
ages during development observed in blood islands
does not mean that their direct progenitors are common
at each blood island level, but that the close association
is needed for their function. One possibility is that be-
cause they eventually form vascular networks by fusion,
the structure of blood islands could be considered as
a preliminary and temporal architecture in preparation
for the future formation of at least the vitelline vascula-
ture, and both progenitors are recruited to the yolk
sac to form blood islands. Another possibility is that,
serving as microenvironment cells, endothelial cells
supply blood cells with growth factors and extracellular
Recently, a model has been proposed that all the pro-
genitors of yolk sac blood islands are originally gener-
ated as a single large cell mass in the yolk sac mem-
brane at the late neural plate stage and are subdivided
by endothelial cell layers later (Ferkowicz and Yoder,
2005). However, from our results, there seem to be sev-
eral separated mesodermal cell masses within the yolk
sac membrane at early neural plate stage embryos and
they grow rapidly and are divided or generated indepen-
dently as the yolk sac membrane expands (Figure S4).
Moreover, our conclusion that there are endothelial
poietic cells based on whole-embryo analysis is not
affected by this model.
Relationship between the Hemangioblast
and Lineage-Committed Progenitors
mangioblasts within gastrulating embryos. The number
of such hemangioblasts detected in one embryo was
in most cases one to five, and they were found between
the mid streak and the neural plate stage (Huber et al.,
2004), which is equivalent to being between E7.0 and
E7.75 (Downs and Davies, 1993). Snow (1977) examined
the cell cycle of cells in early mouse embryos (from E4.5
to E7.5) and estimated the doubling time of mesodermal
cells between E6.5 and E7.5 to be from 13.9 to 22.2 hr,
which is considerably longer than that of ectodermal
and endodermal cells at the same stage; therefore, it is
not likely that the small number of bipotential progeni-
tors within the embryo gives rise to all hematopoietic
together, we suggest from these results that separate
hematopoietic and endothelial progenitors mainly give
rise to the fate their names suggest, and that there are
few true hemangioblasts that generate both lineages.
In fact, our results of whole-embryo analysis strongly
suggest that most endothelial cells are not derived
from a clonal hemangioblast. Moreover, the results of
the Flk1-labeling experiments suggest that the origins
of hematopoietic cells are heterogeneous, many of
them from Flk1-negative precursors, that is, not from
the hemangioblast, while blood vessels are derived
from Flk1-positive progenitors. It isyet to bedetermined
whether the vitelline vessels that connect yolk sac to
embryo allow yolk sac hematopoietic stem cells to
seed the embryo with definitive hematopoietic stem
cell precursors (Weissman et al., 1977, 1978).
A Model of Hematopoietic Development
in the Yolk Sac
Based on our findings and previous findings that show
the presence of hemangioblasts as in vivo entities
(Huber et al., 2004), we can outline three models of
yolk sac hematopoiesis (Figure 6F). In model 1, all of
the endothelial and hematopoietic cells within yolk sac
blood islands are derived from clonal hemangioblasts.
Because we showed in this study that there are endo-
thelial specific progenitors not derived from clonal he-
mangioblasts, we can exclude model 1. In the second
model, the hemangioblasts and lineage-committed pro-
genitors coexist. In the third model, hemangioblasts can
differentiate to only endothelial lineages, to only hema-
topoietic lineages, or to both lineages. In this model,
the hemangioblasts are defined by their potential to
give rise to both lineages, as revealed in vitro. Model 2
and model 3 differ in that there are lineage-committed
progenitors not derived from bipotential hemangio-
blasts in model 2, whereas both lineages are derived
from bipotential progenitors in model 3. Our results
and those of past studies could be consistent with
model 2 or model 3. However, if hemangioblasts are
Flk1 positive, our result that many of the hematopoietic
cells were not labeled by the Flk1-cre system suggests
that the progenitors are heterogeneous.
One potential experiment that could distinguish be-
tween model 2 and model 3 would be that if model 3 is
correct, the fate of progenitors isolated from gastrulat-
ing embryos at the primitive streak stage could be
changed specifically to endothelial or hematopoietic lin-
eages depending on culture conditions. Another clue
might come from studies of adult tissue-specific stem
cells. For example, in the case of adult hematopoietic
stem cells in bone marrow, single-cell transplant exper-
iments into lethally irradiated mice have clearly demon-
strated that all of the blood cell lineages can be gener-
ated from the long-term subset of hematopoietic stem
cells (Smith et al., 1991; Cao et al., 2004; Uchida, 1992;
ficult, if a single hemangioblast can produce all yolk sac
hematopoiesis in the situation where yolk sac hemato-
poiesis and endothelial production is defective, such
as in Flk-1-deficient embryos (Shalaby et al., 1997),
such single cells placed in the early yolk sac should
reveal this potential.
In conclusion, we showed that the origin of yolk sac
blood islands in vivo is polyclonal, and endothelial and
hematopoietic progenitors act mainly in an independent
manner. Our results showed that some, and probably
most, endothelial cells in the yolk sac blood islands
are generated from endothelial specific progenitors
not derived from clonal hemangioblasts, and the model
that all the endothelial and hematopoietic cells in the
yolk sac blood islands are generated from clonal he-
mangioblasts was excluded. Although this study does
not rule out the existence of hemangioblasts that con-
tribute to both lineages, our results suggest greater het-
erogeneity in terms of the origin of yolk sac hematopoi-
esis than previously proposed.
Construction of Knockin Vectors and Loxp-Based
The PacI, AscI, and XbaI sites were created at the XbaI site of the
Rosa26-1 vector (Zambrowicz et al., 1997) and the tk-neo cassette
was subcloned into the XbaI site. The EGFP, ECFP (BD Clontech,
Palo Alto, CA), mRFP1 (Campbell et al., 2002), and LacZ cDNAs
were subcloned into the cloning site of pCAGGS vector (Miyazaki
et al., 1989). The fragment containing the promoter, cDNA, and
polyA/enhancer was cut out as a PacI-AscI fragment and subcloned
into the PacI-AscI sites of the Rosa26-1-tk-neo vector. A loxp-XhoI-
loxp-NheI-EcoRI synthetic oligodeoxynucleotide was subcloned
into the EcoRI site of pCAGGS-puro. The XhoI-SalI stuffer sequence
from pBabe (Morgenstern and Land, 1990) was inserted into the
XhoI site between two loxp sites of the loxp-stop-loxp-pCAGGS-
puro vector. To construct the loxp-stop-loxp-EGFP and loxp-stop-
loxp-ECFP vectors, EcoRI fragments of EGFP and ECFP cDNAs
were subcloned into the EcoRI site of the loxp-stop-loxp-
pCAGGS-puro vector. To construct the loxp-LacZ-loxp-pCAGGS
vector, the LacZ-polyA/enhancer sequences were cut out from
LacZ-pCAGGS vector as an XhoI fragment and subcloned into the
XhoI site between two loxp sites of the loxp-stop-loxp-pCAGGS-
puro vector. To construct the loxp-CDG vector, a loxp-NheI syn-
thetic oligonucleotide was introduced into the HindIII site at the 30
region of the polyA/enhancer sequence of EGFP-pCAGGS, ECFP-
enhancer-loxp sequences were cut out as XbaI-NheI fragments
and subcloned into the NheI site at just after the second loxp site
of the loxp-stop-loxp-pCAGGS-puro vector. This abolished the 50
NheI site but not the 30NheI site. The XbaI-NheI fragment of the sec-
ond fluorescent cDNA-polyA/enhancer was subcloned into the NheI
site, and the XbaI-NheI fragment of the third fluorescent cDNA-
polyA/enhancer was subcloned into the NheI site, respectively.
The constructs were verified by sequencing and by transient trans-
fection into Cos7 cells.
Cell Culture, Transfection, and Screening of Targeted Clones
R1 and Flk1-cre ES cells (provided by Dr. T.N. Sato; Motoike et al.,
2003) were cultured in Dulbecco’s modified Eagle’s medium
(DMEM) (Invitrogen, Carlsbad, CA) containing 15% fetal calf serum
(Hyclone, South Logan, UT), 1.0 3 103U/ml leukemia inhibitory fac-
tor, 1 mM sodium pyruvate (Invitrogen), 2 mM L-glutamine (Invitro-
(Invitrogen), and 1% penicillin/streptomycin (Invitrogen) (ES me-
dium). ES cells were maintained on irradiated mouse embryonic fi-
broblast feeder cells. To generate ES clones, 25 mg of vectors was
linearized with ScaI restriction enzyme and electroporated into R1
cells or Flk1-cre ES cells at the condition of 300 volts, 500 mF by us-
ing Gene Pulser (Bio-Rad, Hercules, CA). Then cells were selected in
the presence of 300 mg/ml of G418 or 1 mg/ml of puromycin, from 18
hr after transfection. Between day 7 and 10 after transfection, resis-
tant colonies were picked up and expanded. Genomic Southern blot
hybridization was performed on DNA from ES clones digested with
HindIII. The 50probe used detects a 6.1 kb wild-type band and an
8.5 kb (for EGFP and ECFP) or a 4.4 kb targeted band (for mRFP1)
duetothe presence ofan extraHindIII site inthe30noncoding region
of the mRFP1 cDNA. Genomic PCR with a sense primer (50-
CCTAAAGAAGAGGCTGTGCTTTGG-30; Rosa26 50region) and an
antisense primer (50-GGGCTATGAACTAATGACCCCG-30; CAG pro-
moter region) was performed by using LA-Taq (Takara Mirus,
Madison, WI) to detect cells carrying the targeted allele according
to the manufacturer’s protocol. Separation of fluorescent-positive
and -negative cells was performed with a FACS ARIA (BD Biosci-
ences, San Jose, CA). For induction of an Mx1 promoter-mediated
cre expression, polyinosinic acid-polycytidylic acid (PIPC) (Sigma,
St. Louis, MO) was added to the medium at the concentration of
2.5 mg/ml and incubated for 24 hr.
Mice and Blastocyst Injection
Mice were bred and maintained at the Stanford University Research
Animal Facility in accordance with Stanford University guidelines.
C57BL/Ka female mice were superovulated with 5 IU of pregnant
mare serum gonadotropin (PMSG) (Chemicon International, Teme-
cula, CA) and 5 IU of human choronic gonadotropin (hCG) (Chemi-
con International) and mated by a standard protocol. Blastocysts
were collected at E3.5. For each blastocyst, a mixture of 15 ES cells
was injected. For single-cell injection, Rosa-EGFP, Rosa-ECFP, and
Rosa-mRFP1 ES cloneswereputon aseparateplaceon aninjection
chamber, and each one of the three clones were picked up and in-
jected into a blastocyst. Injected blastocysts were then transferred
into the uterus of Day 2.5 pseudopregnant B6CBAF1 mice (Jackson
Laboratory, Bar Harbor, ME). Recipient mice were sacrificed and
embryos were taken out and analyzed at the presomite stage.
Embryos were fixed in 4% paraformaldehyde at 4?C. Then embryos
were washed with PBS, cryoprotected overnight in 30% sucrose,
and quick-frozen in optimum cutting temperature (OCT) compound.
Frozen sections (5–7 mm) were cut at 220?C from OCT-embedded
tissues using a microtome (Bright Instruments, Huntingdon, UK).
To detect expression of b-galactosidase, samples were stained
with anti-b-galactosidase antibody (Cappel, West Chester, PA) and
were analyzed by standard fluorescence microscopy using a Nikon
Eclipse E800 immunofluorescence microscope, and by laser scan-
ning confocal microscopy using the LSM 510 confocal laser scan-
ning microscope with a Coherent Mira 900 tunable Ti; sapphire laser
for two-photon excitation (Zeiss, Thornwood, NY). To distinguish
positive fluorescence signals from autofluorescence background
signals, merged images of EGFP, ECFP, and Red channels were
analyzed even when RFP-expressing cells were not injected (Fig-
ure 5). Autofluorescent signals were almost equally emitted to the
three channels and therefore were detected as white signals. When
necessary, counterstaining of the nucleus with Hoechst 33342
(Sigma) at the concentration of 1 mM for 2 min was performed.
For hematopoietic and endothelial cells of each blood island, contri-
bution of fluorescent markers was scored and difference was as-
sessed by one-way analysis of variance (ANOVA). Statistical signif-
icance was accepted within 95% confidence limits.
Supplemental Data include four figures and five tables and are avail-
able at http://www.developmentalcell.com/cgi/content/full/11/4/
Wethank T.N.Sato forFlk1-creES cells,J.MiyazakiforthepCAGGS
vector, P. Soriano for the Rosa26-1 vector, R.Y. Tsien for the mRFP1
cDNA, R. Kuehn for the Mx1-cre vector, L. Hidalgo and D.B. Escoto
for animal care, J. Mulholland and K. Lee for assistance with confo-
cal microscopy, J. Chen for assistance with statistical analysis,
M. Muijtjens for assistance with animal experiments, C. Richter for
assistance with cell culture, C.C. Chen and T. Raveh for technical
advice, S-I. Nishikawa, D. Sugiyama, and H. Zeng for helpful discus-
sions, C. Forsberg for critical reading of the manuscript, and L. Jer-
abek for laboratory management. H.U. was supported in part by the
Postdoctoral Fellowship for Research Abroad, Japan Society for
the Promotion of Science, Uehara Memorial Foundation Fellowship,
the National Cancer Center Research Institute, Japan, and currently
by a gift from the Floren Family Fund. This work was supported
by grants from the USPHS National Institutes of Health to I.L.W.
(2R01 CA86065-06 and 5 P01 DK53074-09). I.L.W. has over
$10,000 in Amgen stock and is a Director of both StemCells, Inc.,
and Cellerant, Inc. We believe the data in this paper have no direct
value to any of these 3 entities.
Origin of Yolk Sac Blood Islands
Received: March 5, 2006
Revised: June 16, 2006
Accepted: August 2, 2006
Published: October 2, 2006
Akashi, K., and Weissman, I.L. (2001). Stem cells and hematolym-
phoid development. In Hematopoiesis: A Developmental Approach,
L.I. Zon, ed. (New York: Oxford University Press), pp. 15–34.
Barker, J.E. (1968). Development of the mouse hematopoietic sys-
tem. I. Types of hemoglobin produced in embryonic yolk sac and
liver. Dev. Biol. 18, 14–29.
Brotherton, T.W., Chui, D.H., Gauldie, J., and Patterson, M. (1979).
Hemoglobin ontogeny during normal mouse fetal development.
Proc. Natl. Acad. Sci. USA 76, 2853–2857.
Campbell, R.E., Tour, O., Palmer, A.E., Steinbach, P.A., Baird, G.S.,
Zacharias, D.A., andTsien,R.Y. (2002). Amonomericredfluorescent
protein. Proc. Natl. Acad. Sci. USA 99, 7877–7882.
Cao, Y.A., Wagers, A.J., Beilhack, A., Dusich, J., Bachmann, M.H.,
Negrin, R.S., Weissman, I.L., and Contag, C.H. (2004). Shifting foci
of hematopoiesis during reconstitution from single stem cells.
Proc. Natl. Acad. Sci. USA 101, 221–226.
Choi, K., Kennedy, M., Kazarov, A., Papadimitriou, J.C., and Keller,
G. (1998). A common precursor for hematopoietic and endothelial
cells. Development 125, 725–732.
Christensen, J.L. (2003). Mechanisms and mediators of hemato-
poietic stem cell fate. PhD thesis, Stanford University, Stanford,
Cumano, A., Dieterlen-Lievre, F., and Godin, I. (1996). Lymphoid po-
tential, probed before circulation in mouse, is restricted to caudal
intraembryonic splanchnopleura. Cell 86, 907–916.
Cumano, A., Ferraz, J.C., Klaine, M., Di Santo, J.P., and Godin, I.
(2001). Intraembryonic, but not yolk sac hematopoietic precursors,
isolated before circulation, provide long-term multilineage reconsti-
tution. Immunity 15, 477–485.
Downs, K.M., and Davies, T. (1993). Staging of gastrulating mouse
embryos by morphological landmarks in the dissecting microscope.
Development 118, 1255–1266.
Drake, C.J., and Fleming, P.A. (2000). Vasculogenesis in the day 6.5
to 9.5 mouse embryo. Blood 95, 1671–1679.
Eichmann, A., Corbel, C., Nataf, V., Vaigot, P., Breant, C., and Le
Douarin, N.M. (1997). Ligand-dependent development of the endo-
thelial and hemopoietic lineages from embryonic mesodermal cells
expressing vascular endothelialgrowth factor receptor2.Proc. Natl.
Acad. Sci. USA 94, 5141–5146.
Ferkowicz, M.J., and Yoder, M.C. (2005). Blood island formation:
longstanding observations and modern interpretations. Exp. Hema-
tol. 33, 1041–1047.
Garcia-Otin, A.L., and Guillou, F. (2006). Mammalian genome target-
ing using site-specific recombinases. Front. Biosci. 11, 1108–1136.
Gardner, R.L., and Rossant, J. (1979). Investigation of the fate of 4-5
day post-coitum mouse inner cell mass cells by blastocyst injection.
J. Embryol. Exp. Morphol. 52, 141–152.
Gardner, R.L., Lyon, M.F., Evans, E.P., and Burtenshaw, M.D. (1985).
Clonal analysis of X-chromosome inactivation and the origin of the
germ line in the mouse embryo. J. Embryol. Exp. Morphol. 88,
Haar, J.L., and Ackerman, G.A. (1971). A phase and electron micro-
scopic study of vasculogenesis and erythropoiesis in the yolk sac of
the mouse. Anat. Rec. 170, 199–223.
Haemangioblast commitment is initiated in the primitive streak of
the mouse embryo. Nature 432, 625–630.
Kabrun, N., Buhring, H.J., Choi, K., Ullrich, A., Risau, W., and Keller,
G. (1997). Flk-1 expression defines a population of early embryonic
hematopoietic precursors. Development 124, 2039–2048.
Kaufman, M.H. (1995). The Atlas of Mouse Development, Second
Edition (San Diego: Academic Press).
Kennedy, M., Firpo, M., Choi, K., Wall, C., Robertson, S., Kabrun, N.,
and Keller, G. (1997). A common precursor for primitive erythropoi-
esis and definitive haematopoiesis. Nature 386, 488–493.
Kinder, S.J., Tsang, T.E., Quinlan, G.A., Hadjantonakis, A.K., Nagy,
A., and Tam, P.P. (1999). The orderly allocation of mesodermal cells
to the extraembryonic structures and the anteroposterior axis dur-
ing gastrulation of the mouse embryo. Development 126, 4691–
Kos, C.H. (2004). Cre/loxP system for generating tissue-specific
knockout mouse models. Nutr. Rev. 62, 243–246.
Lawson, K.A., and Pedersen, R.A. (1992). Early mesoderm formation
in the mouse embryo. In Formation and Differentiation of Early Em-
bryonic Mesoderm, R. Bellairs, E.J. Sanders, and J.W. Lash, eds.
(New York: New York Plenum Press), pp. 33–46, NATO ASI Series
Lawson, K.A.,Meneses, J.J., andPedersen,R.A. (1991). Clonal anal-
ysis of epiblast fate during germ layer formation in the mouse em-
bryo. Development 113, 891–911.
McGrath, K.E., Koniski, A.D., Malik, J., and Palis, J. (2003). Circula-
tion is established in a stepwise pattern in the mammalian embryo.
Blood 101, 1669–1676.
Medvinsky, A., and Dzierzak, E. (1996). Definitive hematopoiesis is
autonomously initiated by the AGM region. Cell 86, 897–906.
Mintz, B., and Silvers, W.K. (1967). ‘‘Intrinsic’’ immunological toler-
ance in allophenic mice. Science 158, 1484–1486.
Miyazaki, J., Takaki, S., Araki, K., Tashiro, F., Tominaga, A., Takatsu,
K., and Yamamura, K. (1989). Expression vector system based on
the chicken b-actin promoter directs efficient production of interleu-
kin-5. Gene 79, 269–277.
Morgenstern, J.P., and Land, H. (1990). Advanced mammalian gene
transfer: high titre retroviral vectors with multiple drug selection
markers and a complementary helper-free packaging cell line. Nu-
cleic Acids Res. 18, 3587–3596.
Morrison, S.J., and Weissman, I.L. (1994). The long-term repopulat-
ing subset of hematopoietic stem cells is deterministic and isolat-
able by phenotype. Immunity 1, 661–673.
Motoike, T., Markham, D.W., Rossant, J., and Sato, T.N. (2003). Ev-
idence for novel fate of Flk1+ progenitor: contribution to muscle lin-
eage. Genesis 35, 153–159.
Murray, P.D.F. (1932). The development in vitro of the blood of the
early chick embryo. Proc. R. Soc. Lond. B Biol. Sci. 111, 497–521.
Palis, J., Robertson, S., Kennedy, M., Wall, C., and Keller, G. (1999).
Development of erythroid and myeloid progenitors in the yolk sac
and embryo proper of the mouse. Development 126, 5073–5084.
Robertson, S., Kennedy, M., and Keller, G. (1999). Hematopoietic
commitment during embryogenesis. Ann. N Y Acad. Sci. 872, 9–16.
Rossant, J., Gardner, R.L., and Alexandre, H.L. (1978). Investigation
of the potency of cells from the postimplantation mouse embryo by
blastocyst injection: a preliminary report. J. Embryol. Exp. Morphol.
Sasaki, K., and Kendall, M.D. (1985). The morphology of the haemo-
olar changes. J. Anat. 140, 279–295.
Serbedzija, G.N., Bronner-Fraser, M., and Fraser, S.E. (1989). A vital
dye analysis of the timing and pathways of avian trunk neural crest
cell migration. Development 106, 809–816.
Shalaby, F., Rossant, J., Yamaguchi, T.P., Gertsenstein, M., Wu,
X.F., Breitman, M.L., and Schuh, A.C. (1995). Failure of blood-island
formation and vasculogenesis in Flk-1-deficient mice. Nature 376,
Shalaby, F., Ho, J., Stanford, W.L., Fischer, K.D., Schuh, A.C.,
Schwartz, L., Bernstein, A., and Rossant, J. (1997). A requirement
for Flk1 in primitive and definitive hematopoiesis and vasculogene-
sis. Cell 89, 981–990.
Smith, L.G., Weissman, I.L., and Heimfeld, S. (1991). Clonal analysis
of hematopoietic stem-cell differentiation in vivo. Proc. Natl. Acad.
Sci. USA 88, 2788–2792.
Snow, M.H.L. (1977). Gastrulation in the mouse: growth and region-
alization of the epiblast. J. Embryol. Exp. Morphol. 42, 293–303.
Tam, P.P., and Beddington, R.S. (1987). The formation of mesoder-
mal tissues in the mouse embryo during gastrulation and early or-
ganogenesis. Development 99, 109–126.
Uchida, N. (1992). Characterization of mouse hematopoietic stem
cells. PhD thesis, Stanford University, Stanford, California.
Weissman, I.L., Baird, S., Gardner, R.L., Papaioannou, V.E., and
Raschke, W. (1977). Normal and neoplastic maturation of T-lineage
lymphocytes. Cold Spring Harb. Symp. Quant. Biol. 41, 9–21.
Weissman, I., Papaioannou, V., and Gardner, R. (1978). Fetal hema-
topoietic origins of the adult hematolymphoid system. In Differenti-
ation of Normal and Neoplastic Hematopoietic Cells, B. Clarkson,
P.A. Marks, and J.E. Till, eds. (New York: Cold Spring Harbor Labo-
ratory Press), pp. 33–47.
Zambrowicz, B.P., Imamoto, A., Fiering, S., Herzenberg, L.A., Kerr,
W.G., and Soriano, P. (1997). Disruption of overlapping transcripts
in the ROSA b geo 26 gene trap strain leads to widespread expres-
sion of b-galactosidase in mouse embryos and hematopoietic cells.
Proc. Natl. Acad. Sci. USA 94, 3789–3794.
Zong, H., Espinosa, J.S., Su, H.H., Muzumdar, M.D., and Luo, L.
(2005). Mosaic analysis with double markers in mice. Cell 121,
Origin of Yolk Sac Blood Islands