ArticlePDF Available

Wnt-3a regulates somite and tailbud formation in the mouse embryo


Abstract and Figures

Amphibian studies have implicated Wnt signaling in the regulation of mesoderm formation, although direct evidence is lacking. We have characterized the expression of 12 mammalian Wnt-genes, identifying three that are expressed during gastrulation. Only one of these, Wnt-3a, is expressed extensively in cells fated to give rise to embryonic mesoderm, at egg cylinder stages. A likely null allele of Wnt-3a was generated by gene targeting. All Wnt-3a-/Wnt-3a- embryos lack caudal somites, have a disrupted notochord, and fail to form a tailbud. Thus, Wnt-3a may regulate dorsal (somitic) mesoderm fate and is required, by late primitive steak stages, for generation of all new embryonic mesoderm. Wnt-3a is also expressed in the dorsal CNS. Mutant embryos show CNS dysmorphology and ectopic expression of a dorsal CNS marker. We suggest that dysmorphology is secondary to the mesodermal and axial defects and that dorsal patterning of the CNS may be regulated by inductive signals arising from surface ectoderm.
Content may be subject to copyright.
Wnt-3a regulates somite and tailbud
formation in the mouse embryo
Shinji Takada/ Kevin L. Stark,^ Martin J. Shea,^ Galya Vassileva, Jill A. McMahon/
and Andrew P. McMahon*
Roche Institute of Molecular Biology, Roche Research Center, Nutley, New Jersey 07110 USA
Amphibian studies have implicated Wnt signaling in the regulation of mesoderm formation, although direct
evidence is lacking. We have characterized the expression of 12 mammalian Wnt-genes, identifying three that
are expressed during gastrulation. Only one of these, Wnt-3a, is expressed extensively in cells fated to give rise
to embryonic mesoderm, at egg cylinder stages. A likely null allele of Wnt-3a was generated by gene targeting.
All Wiit-3fl~/Wnt-3a~ embryos lack caudal somites, have a disrupted notochord, and fail to form a tailbud.
Wnt-Sa may regulate dorsal (somitic) mesoderm fate and is required, by late primitive steak stages, for
generation of all new embryonic mesoderm. Wnt-3a is also expressed in the dorsal CNS. Mutant embryos
show CNS dysmorphology and ectopic expression of a dorsal CNS marker. We suggest that dysmorphology is
secondary to the mesodermal and axial defects and that dorsal patterning of the CNS may be regulated by
inductive signals arising from surface ectoderm.
[Key Words: Wnt; gastrulation; mesoderm formation; somite; tailbud; gene targeting]
Received November 9, 1993; revised version accepted December 8, 1993.
Cell-cell interaction plays an important role in the de-
velopment of all organisms. In the vertebrate, consider-
able progress has been made in recent years in identify-
ing peptide growth factors, which may mediate the early
inductive interactions that result in patterning of the
embryo (for review, see Jessell and Melton 1992; Slack
and Tannahill 1992; McMahon 1993). Most of these
studies have focused on the amphibian embryo, where
the ease of experimental manipulation has led to impor-
tant advances in our understanding of early inductive
interactions, particularly those governing the generation
of mesodermal components at gastrulation (for review,
see Slack and Tannahill 1992; Sive 1993).
Mesoderm is formed as an annulus around the equator
of the amphibian embryo in response to vegetal-derived,
maternally encoded signals. Members of two growth fac-
tor families, fibroblast growth factor (FGF) and trans-
forming growth factor-p (TGF-p), are capable of inducing
mesoderm formation in isolated explants of blastula an-
imal caps, cells that do not normally generate mesoderm
in the embryo. Thus, these factors are implicated in nor-
mal mesoderm induction in vivo. This conclusion is
strengthened by the observed inhibition of mesoderm
formation in embryos expressing dominant-negative
forms of FGF (Amaya et al. 1991) and activin (a member
of the TGF-p superfamily; Hemmati-Brivanlou and Mel-
Ptesent addresses: ^Department of Cellular and Developmental Biology,
The Biological Laboratories, Harvard University, Cambridge, Massachu-
setts 02138 USA; ^Amgen, Inc., Thousand Oaks, California 91320 USA;
^Institute for Molecular Genetics, Baylor College of Medicine, Houston,
Texas 77030 USA.
ton 1992) receptors. Interestingly, mesoderm induction
by FGFs and TGF-p-related factors is qualitatively differ-
ent. FGF induces mostly ventral cell types, such as me-
sothelium and blood cells (Slack et al. 1987; Kimelman
et al. 1988), whereas activin and
induce dorsal me-
soderm of the notochord (Smith 1987; Smith et al. 1990;
Sokol et al. 1990; Thompson and Melton 1993). All fac-
tors will induce somitic mesoderm.
Finally, inductive signals released during gastrulation
by cells of the dorsal lip of the blastopore, the so-called
Organizer region, are required to generate the full com-
plement of dorsal mesodermal cell types. Thus, induc-
tion and patteming of the embryonic mesoderm proba-
bly result from the combined action of multiple signals,
some of which appear to be localized in the cleaving and
gastrulating embryo.
A third family of putative signaling molecules, the
Wnt family, has also been implicated in embryonic me-
soderm induction. Wnt genes are widely dispersed in in-
vertebrate and vertebrate organisms. There are at least
15 vertebrate members; all encode likely secreted pro-
many of which are expressed during early verte-
brate development (for review, see McMahon 1992;
Nusse and Varmus 1992). Injection of RNAs encoding
Wnt-1 or Wnt-8 into one- or two-cell amphibian eggs
results in axis duplication or rescue of the entire axis in
UV-irradiated, axial-deficient embryos (McMahon and
Moon 1989; Smith and Harland 1991; Sokol et al. 1991).
Both of these members are capable of inducing a new
Organizer region or of acting as an Organizer-derived,
dorsalizing signal, which either directly induces dorsal
174 GENES & DEVELOPMENT 8:174-189 © 1994 by Cold Spring Harbor Laboratory Press ISSN 0890-9369/94 $5.00
Cold Spring Harbor Laboratory Press on October 4, 2016 - Published by genesdev.cshlp.orgDownloaded from
Regulation of gastiulation by WnU3a
mesoderm (Sokol et al. 1991) or, more likely, acts to
modify the response of receiving cells to other peptide
growth factors (Christian et al. 1992). However, although
Wnt-1 and Wnt-8
implicated in patterning of dorsal
mesoderm, expression of these signals is inconsistent
with this view.
Wnt-1 expression is restricted to the later arising cen-
tral nervous system (CNS), where it regulates midbrain
development (McMahon and Bradley 1990; Thomas and
Capecchi 1990; Thomas et al. 1991; McMahon et al.
In contrast, Wnt-8 is expressed during gastrulation
but exclusively in ventral regions (Christian et al. 1991;
Smith and Harland 1991). Moreover, recent experiments
ectopically expressing Wnt-8, during gastrulation, sug-
gest that its real function may actually lie in restricting
induction of dorsal mesoderm (Christian and Moon
There are several explanations to these apparently
conflicting data. For example, ectopically expressed
Wnt-1 and Wnt-8 may mimic the action of an endoge-
nous Wnt ligand by interacting with the ligand's recep-
tor. This would suggest that some, as yet unidentified,
Wnt member does regulate dorsal mesoderm formation.
Alternatively, Wnt-1 and Wnt-8 may trigger a receptor-
mediated pathway that is not normally stimulated at
these early embryonic stages. If such a receptor couples
intracellularly to the bona fide dorsal mesoderm-induc-
ing signaling pathway, this may potentially explain the
observed results. As Wnt receptors have not been iden-
tified, this issue cannot be directly resolved. However, a
better understanding of the spatial and temporal expres-
sion of Wnt signals that may regulate gastrulation is
likely to be informative.
In the mouse, gastrulation is initiated at 6.5 days post-
coitum (dpc) by the delamination of epiblast cells into
the primitive streak at the posterior limit of the embryo
(for review, see Beddington 1983). By mid-somite stages
(9.25 to 9.5 dpc in the mouse), the primitive streak is
replaced by the tailbud, which consists of an aggregate of
mesodermal precursors at the caudal extremity of the
embryo. New mesoderm continues to arise from the tail-
bud for several days (Schoenwolf 1977). Thus, the gener-
ation and patterning of mesodermal cell types along the
embryonic axis occurs over a relatively long period of
To determine whether Wnt signals participate in gas-
trulation in the mouse, we examined the expression of
12 members of the Wnt gene family. Three members,
Wnt-3a, Wnt-Sa, and Wnt-Sb, are expressed in discrete
spatial domains in the primitive streak, and expression
persists into the tailbud. Expression of Wnt-5a and Wnt-
Sb is largely restricted to extraembryonic mesoderm pre-
cursors from early to late egg cylinder stages. In contrast,
Wnt-3a is also expressed in cells that generate the
somites and lateral mesoderm, suggesting that Wnt-3a
signaling may normally regulate the formation of these
mesodermal cell types. To address the function of Wnt-
Sa, we generated a likely null allele by gene targeting.
Analysis of Wnt-3a homozygous mutant embryos indi-
cates that Wnt-3a plays a key role in the regulation of
mammalian gastrulation.
Wnt gene expression during mouse gastrulation
To analyze the expression of Wnt genes during mouse
gastrulation, we performed in situ hybridization with
probes specific to Wnt-3, Wnt-3a, Wnt-4, Wnt-5a, Wnt-
Wnt-6, Wnt-1 a, Wnt-lb, Wnt-10, and Wnt-11. Wnt-1
and Wnt-2 expression was analyzed in earlier studies
(Wilkinson et al. 1987a; McMahon and McMahon 1989).
Only Wnt-3a, Wnt-Sa, and Wnt-Sb are expressed in the
primitive streak and tailbud (data not shown). To obtain
a comprehensive understanding of the temporal and spa-
tial pattem of these three members, we performed serial
section in situ hybridization to egg cylinder-stage em-
bryos and whole-mount in situ hybridization to late
streak and early tailbud, somite-stage embryos. At the
egg cylinder stage, Wnt expression patterns were mapped
relative to expression of the Brachyury (T) gene, one of
the first genes to be activated at the onset of gastrulation
(Herrmann 1991). Brachyury is implicated in mesoder-
mal specification, possibly as a direct response to meso-
derm-inducing factors (Cunliffe and Smith 1992), and is
essential for development of the notochord and caudal
axis (for review, see Herrmann 1992). Thus, Brachyury
expression, which first occurs in both epiblast and me-
soderm within the streak and then later in the head pro-
notochord, node, and tailbud, is an excellent tem-
poral and spatial marker of gastrulation and, more spe-
cifically, of the formation of dorsal midline structures
(Wilkinson et al. 1990; Herrmann 1991).
At early primitive streak stages in the mouse, when
the T gene is first expressed (6.5 dpc), Wnt-Sa and Wnt-
Sb transcripts are detected—Wnt-5fl, only in mesoder-
mal cells, and Wnt-Sb, in epiblast and mesoderm within
the streak (Fig. 1). It is not clear whether cells, at this or
any later stage, coexpress Wnt-Sa and Wnt-Sb as the in
situ technique does not allow single-cell resolution. By
7.0 dpc, the primitive streak has elongated distally, and
expression of the T gene occupies the proximodistal ex-
tent of the streak but is restricted medially within the
streak to epiblast and mesoderm cells (Fig. 1). Wnt-Sa
and Wnt-Sb remain posteriorly restricted. In transverse
section, Wnt-Sb expression is restricted to medial epi-
blast and mesoderm within the streak, whereas Wnt-Sa
expression is absent in epiblast and extends more later-
ally in the mesoderm with a broader distribution than T
gene expression (Fig. 1). Parasagittal sections at 7.0 dpc
only detect laterally distributed Wnt-Sa expression (Fig.
By the extended-streak stages (7.5 dpc), the third Wnt
member, Wnt-3a, is widely expressed within the streak.
Hybridization is patchy, suggesting that not all cells ex-
press Wnt-3a, but expression extends along much of the
length of the streak, its distal limit lying just anterior to
the node (Fig. 1). Brachyury expression resembles that of
Wnt-3a in the streak but extends anteriorly into the node
and midline mesoderm of the head process (Fig. 1). Ex-
pression of Wnt-Sa and Wnt-Sb remains posteriorly lo-
calized up to 7.75 dpc, extending into newly formed ex-
traembryonic mesoderm of the allantois (Fig. 1). In sum-
Cold Spring Harbor Laboratory Press on October 4, 2016 - Published by genesdev.cshlp.orgDownloaded from
Takada et al.
Wnt-3a Wnt-5a Wnt-5b
Figure 1. Expression of Brachyuiy (T gene), Wnt-Sa, Wnt-5a, and Wnt-5h at egg cylinder stages of mouse gastrulation. Adjacent
sagittal (6.5, 7.5, and 7.75 dpc), parasagittal (7.0 dpc), and transverse (7.0 dpc) sections through the primitive streak region were
hybridized w^ith RNA probes specific to each gene. The approximate limits of the primitive streak are marked by the solid lines. The
orientation of egg cylinder sections are as follows: (A) anterior; (Di) distal; (L) lateral; (P) posterior; (Pr) proximal. For a detailed
description, see text. (AC) Amniotic cavity; (AL) allantois; (EP) epiblast; (HP) head process; (M) mesoderm; (N) node; (NP) neural plate.
Initiation of Wnt-Sa expression at 7.0 dpc is indicated by the arrow. The apparent signal within the decidual tissue at several stages
is not due to silver grains but is an artifact of tissue birefringence.
mary, expression of Wnt-Sa and Wnt-Sh from early to
late egg cylinder stages is localized to posterior regions of
the streak that cell fate studies indicate form extraem-
bryonic mesoderm (see Discussion). In contrast, Wnt-Sa
expression extends more anteriorly, suggesting that
Wnt-3a may regulate the formation of embryonic meso-
derm that is fated to arise from the anterior primitive
streak (see Discussion).
Expression of all three Wnt genes continues in the late
streak and tailbud of somite-stage embryos (Fig. 2). At
Cold Spring Harbor Laboratory Press on October 4, 2016 - Published by genesdev.cshlp.orgDownloaded from
Wnt-3a Wnt-5a
Regulation of gastrulation by Wnt-3a
Figure 2. Expression of Wnt-3a, Wnt-Sa, and Wnt-Sb at somite stages. Whole-mount in situ hybridization with digoxygenin-labeled
RNA probes specific to each gene. At 8.5 dpc (10-11 somites), all three Wnt genes are expressed dorsally in the primitive streak region;
however, Wnt-Sa, unlike Wnt-5a and Wnt-5b, is not expressed ventrally (large arrowhead). Wnt-3a expression is initiated in the dorsal
diencephalon (small arrowhead). At 9.5 dpc (22-26 somites) and 9.75 dpc (32-35 somites), Wnt-3a and Wnt-5b are localized exclusively
to the tailbud (TB) region, whereas Wnt-5a expression extends more anteriorly in dorsal and ventral cells. At these stages, Wnt-3a
expression extends ventrolaterally in a specific domain of the diencephalon (large arrowhead) and continues caudally along the entire
neuraxis and rostrally into the dorsal telencephalon (small arrowhead).
early somite stages, Wnt-3a is expressed dorsally and
caudally within the primitive streak, v\^hereas Wnt-5a
and Wnt-5b expression extends rostral and ventral. As
the primitive streak is lost and the tailbud forms, Wnt-
Sa- and W22£-5b-expressing cells are localized to the cau-
dal-most tip of the tailbud, whereas Wnt-5a is expressed
in a broader region. Thus, as the embryo undergoes a
transition from primitive streak to tailbud, expression of
Wnt-3a, Wnt-5a, and Wnt-5b is maintained in the region
that continues to generate new mesoderm. In addition.
Wnt-3a is also expressed in the developing CNS (see be-
Targeting the Wnt-3a gene
Of the three mouse Wnt genes, only Wnt-3a shows a
clear association with the formation of embryonic me-
soderm at egg cylinder stages. To address the function of
Wnt-3a, we generated a likely null allele by gene target-
ing in mouse embryo stem (ES) cells. Figure 3A indicates
Cold Spring Harbor Laboratory Press on October 4, 2016 - Published by genesdev.cshlp.orgDownloaded from
Takada et al.
Expected Restriction Fragments (kb.)
5' probe 3' probe
Wnt-3A gene __//
targeting vector
mutated allele //
(S) N E(S)
' C
1 INE6|
1 1
1 4 ^ Hind
9.3 9.U
1 1
1 1
10.9 6.6
B 5' probe
Hind III Nco I
ES 8A 17A ES 8A 17A
3' probe
ES 8A 17A
Gene targeting at the Wnt-3a lo-
[A] Schematic representation of the
expected gene replacement at the Wnt-3a
The predicted size of diagnostic re-
striction fragments from the wild-type and
targeted Wnt-3a alleles with 5' and 3' spe-
cific genomic probes are indicated, as well
as the region used to initially identify ho-
mologous recombinant clones by PCR. (B)
Southern blot analysis of the targeted 8A
and 17A J7 ES cell lines. Both 5' and 3'
probes detect restriction fragments indica-
tive of the predicted gene replacement
event at the Wnt-Sa locus. (C) Southern
blot analysis of the genotype of 9.5-dpc em-
bryos from heterozygous intercrosses be-
tween Wnt-3a~/
mice. Embryos were
scored for normal (N) or mutant (M) phe-
notypes. The mutant phenotype, loss of
somites and axial truncation, was only ob-
served in homozygous Wnt-3a~/Wnt-3a'
+/+ +/+
the Wnt-3a targeting strategy. The PGKneobpA gene (So-
riano et al. 1991) was inserted into a Smal site (amino
acid position 106; Roelink and Nusse, 1991) in the third
exon of Wnt-3a and the selection cassette flanked by 0.8
kb of 5'- and 4.7 kb of 3'-genomic homology. Thus, a
successful gene replacement at the Wnt-3a locus should
result in truncation of the normal Wnt-3a open reading
frame, producing a small protein product approximately
one-third of the normal size. As all similarly truncated
alleles of Wnt-1 (McMahon and Moon 1989) and the
Drosophila counterpart wingless (A. Martinez-Arias,
comm.) are invariably inactive, w^e expect that suc-
cessful targeting vy^ill generate a null allele of Wnt-3a.
The targeting construct was transfected into the ]7 ES
cell line (P. Swiatek, F. Franco del Amo, and T. Gridley,
in prep.), and homologous recombinants were detected
by PCR (data not shown). Of 192 G418- and FIAU-resis-
tant clones, 2 gave the predicted PCR amplification prod-
uct. These clones, 8A and 17A, were analyzed by South-
ern analysis using 5'- and 3'-specific probes, which flank
Cold Spring Harbor Laboratory Press on October 4, 2016 - Published by genesdev.cshlp.orgDownloaded from
Regulation of gastnilation by
the area of homology in the targeting vector (Fig. 3A).
Hindlll and Ncol digests of genomic DNA give predicted
10.9-kb and 6.6-kb hybridizing bands, respectively, with
the 5'-flanking probe (Fig. 3B), indicative of correct Wnt-
3a targeting. Similarly, an expected 7.5-kb EcoBl frag-
ment v^ras detected in both targeted ES cell lines using
the 3' probe (Fig. 3B). Thus, both clones exhibited the
expected gene replacement at the Wnt-3a locus. Hybrid-
ization with a neo probe failed to detect any additional
sites of integration (data not shown).
cells were used to generate chimeras, sev-
eral of which transmitted the mutated alleles through
the germ line. Adult mice heterozygous for targeted al-
leles were intercrossed to generate homozygous
spring, and litters were genotyped between 1 and 3
weeks after birth by Southern blot analysis. No homozy-
gous offspring were recovered (Table 1), and no signifi-
cant early postnatal lethality was observed. Thus, loss of
Wnt-3a function results in embryonic lethality.
To determine when embryonic lethality occurred and
to examine Wnt-Sa homozygous mutant embryos, litters
were taken at 9.5 to 12.5 dpc. Yolk sac DNA from em-
bryos was examined by Southern blot analysis for wild-
type or targeted alleles (Fig. 3C). Approximately 25% of
the embryos recovered at 9.5 and 10.5 dpc were homozy-
gous for the targeted alleles, and 85% of these were via-
ble (Table 1). However, only 17% of the homozygous
mutant embryos were recovered alive at 12.5 dpc (Table
Thus, loss of Wnt-3a function leads to embryonic
lethality between 10.5 and 12.5 dpc.
Axial truncation in Wnt-3a mutants
Wnt-3a homozygous mutant embryos from both the 8A
and 17A lines gave identical phenotypes that were
readily apparent at 9.5 dpc. Anterior development ap-
peared normal; however, caudal to the forelimbs, the
Table 1. Embryonic and postnatal survival of progeny from
heterozygous intercrosses
+ / +
"Embryos were genotyped at 9.5-12.5 dpc and neonates at 1-3
weeks after
pooled from lines derived from 8A
17A ES
No differences in survival
were observed between these two
Embryos were scored as
alive (A) or dead (D) at the time of collection.
^'Figures in parentheses represent the percentage of
Wnt-Sa" embryos that were alive at each collection stage.
trunk was shortened, the somites were disrupted or
missing, and the neural tube was highly kinked (Fig. 4B).
All homozygous mutant embryos displayed this pheno-
Thus, the observed axial truncation is completely
penetrant. By 12.5 dpc, homozygous mutant embryos,
which have little or no caudal development posterior to
the forelimb, show massive shortening of the axis rela-
tive to wild-type or heterozygous embryos (Fig. 4F). In-
terestingly, two mutant embryos had a single midline
hindlimb at the caudal extremity of the truncated axis
(data not shown). This suggests that in at least some
embryos the hindlimb fields are formed but presumably
fuse at the center of the truncated axis to give a single
limb in the normal position of the tail. Despite the se-
vere axial truncation, mutant embryos do establish a
chorioallantoic placenta with a good embryonic circula-
tion at 9.5 dpc. Thus, formation of the yolk sac meso-
derm and allantois does not appear to require Wnt-3a.
However, it is quite likely that axial truncation leads to
circulatory complications, and this factor probably leads
to embryonic death at later stages.
Dorsal mesoderm development is defective in Wnt-3a
The axial truncations observed in Wnt-3a homozygous
mutants suggest that gastrulation is severely disrupted.
To examine the phenotype in more detail we performed
histological analysis at 9.5 dpc. Rostral to the forelimbs,
Wnt-3a mutant embryos show a normal organization of
mesodermal derivatives. The notochord is present at the
midline (data not shown), and the first seven to nine
somites appear to be developing normally generating der-
matome, myotome, and sclerotome derivatives (Fig.
In contrast, at the level of the forelimb, somites
appear to be entirely absent (Fig. 5C,D). The neural tube
is folded and kinked leading to multiple cross-sections
through the spinal cord at this axial level (Fig. 5C,D).
Notochord is present at the level of the forelimbs (Fig.
5C,D) but is absent in more posterior regions just caudal
to the forelimb buds (data not shown). Finally, in the
caudal-most region of the mutant axis there is extensive
cell death, which is largely restricted to the dorsal-most
mesodermal cells, and no visible tailbud formation (data
not shown). These results suggest that absence of Wnt-
3a primarily affects development of the somites but has
additional consequences for notochord and tailbud for-
To determine exactly which cell types are present in
mutant embryos and how these are organized along the
body axis, we performed whole-mount in situ hybridiza-
tion with a number of probes to identify specific meso-
dermal populations. At 9.5 dpc, expression of the homeo
box-containing Mox-1 gene is restricted to presomitic
mesoderm and developing somites (Candia et al. 1992).
In Wnt-3a mutants, no Mox-1 expression is seen caudal
to somites 7-9 (Fig. 6A,B), even when embryos were
completely cleared in benzyl alcohol/benzyl benzoate
(Davis et al. 1991; data not shown). Thus, Wnt-3a is not
required for development of somites up to the level of
Cold Spring Harbor Laboratory Press on October 4, 2016 - Published by genesdev.cshlp.orgDownloaded from
Takada et al.
*/ or •/--/-
Figure 4. Phenotype of Wnt-3a /Wnt-Sa embryos at 9.5 and
12.5 dpc. [A-D] 9.5-dpc embryos; {A,B) bright-field and
dark-field views. (£,F) Dark-field view of 12.5-dpc embryos. At
9.5 dpc, formation of the first seven to nine somites
up to the
level of the forelimb bud
is normal; however, no somites are
visible from the forelimbs caudal (arrowhead). In wild-type and
heterozygous embryos, the tailbud (TB) is forming at this stage.
No tailbud is visible in the homozygous mutant embryo (arrow),
and the axis in truncated. At 12.5 dpc (F), caudal development
has ceased at the level of the umbilicus (U) just anterior to
where the hindlimbs (HL) normally form. Wild-type embryos
show extensive tail (T) development, whereas mutant embryos
occasionally have a small degenerate structure (arrowhead) just
caudal to the forelimbs. Diencephalic (D) development appears
normal in homozygous mutant embryos; (AL) allantois; (G) gut;
(L) liver.
the forelimb bud but is required for development of more
posterior somites. In contrast to the dorsally located
somites, ventral (lateral) mesoderm expressing the
Wnt-2 gene, which extends from the heart to the allan-
tois (McMahon and McMahon 1989; Fig. 6C), is present
in Wnt-Sa mutant embryos (Fig. 6D). Thus, Wnt-3a is
apparently not required at similar axial levels for gener-
ation of more ventral mesodermal cell types.
To address notochord and tailbud development in
Wnt-3a homozygous mutants, expression of the
Brachyury and Wnt-5a genes
examined. Brachyury
expression, w^hich is indistinguishable in embryos that
are wild-type or heterozygous for the targeted Wnt-3a
allele, is localized to the notochord along much of the
body axis (Fig. 6E). In contrast, in Wnt-3a homozygous
mutants, notochord expression is disrupted caudal to the
forelimb buds (Fig. 6F). Although a clear notochord is not
seen in the posterior region of Wnt-3a mutants, scattered
Brachyury-expressing cells are visible, usually in two
streams on either side of the midline, suggesting that
notochordal cells are formed but not organized into a
midline structure (Fig. 6F). In addition, we obtained no
evidence for tailbud formation by examining Brachyury
(Fig. 6F), Wnt-5a (Fig. 6G,H), or Wnt-3a (data not shown)
expression. Taken together, our results suggest that
Wnt-3a is initially required at late egg cylinder stages for
somite development, but by late primitive streak/tail-
bud stages, formation of all mesodermal precursors is
dependent, directly or indirectly, on Wnt-3a signaling.
The resulting cessation of gastrulation results in trunca-
tion of the axis.
Disruption of CNS morphogenesis in Wnt-3a mutants
From early somite stages, Wnt-3a is expressed in the
developing CNS. Expression is initiated in the dienceph-
alon and spinal cord (McMahon et al. 1992; Parr et al.
By 9.5-9.75 dpc, Wr2t-3fl-expressing cells extend
along the dorsal midline of the neural tube from the tel-
encephalon to the base of the developing spinal cord.
(Fig. 2; Parr et al. 1993). Moreover, expression extends
ventrally in the diencephalon in a localized domain that
may correspond to a diencephalic segment (Fig. 2; Parr et
1993). Thus, Wnt-3a signaling is implicated in regu-
lating both dorsoventral patterning and forebrain seg-
mentation in the mammalian CNS. To determine
whether either of these processes is affected by the loss
of Wnt-3a, we examined CNS development in Wnt-3a
The neural tube anterior to the forelimbs was morpho-
logically and histologically indistinguishable among ho-
mozygous, heterozygous, and wild-type embryos (Figs. 4
and 5). Moreover, examination of Wnt-3a expression at
9.75 dpc failed to detect any differences in the normal
anterior domain of Wnt-3a expression in the CNS (Fig.
However, from the forelimbs caudal, CNS mor-
phogenesis was highly abnormal. The spinal cord was
kinked, and dorsal fusion of the neural tube was fre-
quently absent. The phenotype is readily apparent in
transverse sections at the level of the forelimbs, where
Cold Spring Harbor Laboratory Press on October 4, 2016 - Published by genesdev.cshlp.orgDownloaded from
Regulation of gastrulation
by Wnt-3a
Figure 5. Histological sections through
9.5-dpc embryos heterozygous and ho-
mozygous for a targeted Wnt-Sa allele.
In horizontal sections anterior to the
forelimb, spinal cord (SC) and somite (ar-
rowhead) development appears normal. At
this axial level somites are differentiating
into dermatomal (DM), myotomal (MY),
and sclerotomal (SL) derivatives. {C,D)
Transverse sections at the level of the fore-
show an absence of somite devel-
opment in mutant (- / -) embryos and se-
vere dysmorphology of the spinal cord. The
clump of cells adjacent to the upper spinal
cord section are probably migrating neural
crest cells (small arrowhead). The noto-
chord is present at the forelimb level (large
arrowhead). (DA) Dorsal aorta; (G) gut.
multiple spinal cord sections can be seen, due in part to
kinking and folding of the main neuraxis under more
caudal regions but also to the apparent pinching off of
small isolated patches of neural tissue (Fig. 5D).
To address dorsoventral patterning, we examined ex-
pression of three dorsally restricted genes, Pax-3,
and Wnt-3a. Pax-3 is expressed in approximately the dor-
sal half of the neural tube from the diencephalon to the
tailbud at 9.5 dpc, as well as in the somites (Fig. 7C,E). In
Wnt-3a~/Wnt-3a~ embryos, Pax-3 expression shows a
similar dorsal restriction extending to the end of the
trunk region (Fig. 7D). However, normal dorsomedial
Pax-3 expression is perturbed in the kinked and open
neural plate, commencing at the position where somites
are first absent (Fig. 7D,F). Moreover, in regions where
the notochord is severely disrupted or missing, Pax-3
expression extends ventrally (Fig. 7D,F).
which is normally expressed in a similar if not
identical domain to Wnt-3a along much of the dorsal
with the exception of the telencephalon and me-
tencephalon (Fig. 7G), is also correctly expressed at the
dorsal aspects forelimb level of the neural tube in Wnt-
3a mutant embryos, up to the forelimb level. Like Pax-3,
Wnt-1 expression extends in the open neural plate to the
caudal limit of the embryo and then projects rostrally as
the neural tube turns through 180° at the end of the
truncated axis (Fig. 7H). Patches of separated neuroepi-
thelium continue to express
Similar results were
obtained examining Wnt-3a expression in Wnt-3a mu-
tant embryos (data not shown).
Wnt-1 expression was examined more closely in serial
sections through Wnt-3a~/Wnt-3a~ embryos caudal to
the forelimbs (Fig. 8). Initially, three sites of Wnt-1 ex-
pression are seen: two dorsal areas corresponding to the
open neural plate and one site corresponding to the dor-
sal region of a more rostral area of the CNS that has
folded under the open neural plate in the caudal part of
the mutant embryo. Interestingly, in progressively more
caudal sections through this embryo, a fourth site of
Wnt-1 expression appears 180° opposite the existing site
in the underfolded spinal cord (Fig. 8B-D). Ectopic ex-
pression of Wnt-1 correlates with the position of the neu-
ral tube in relation to the surface ectoderm. When sepa-
rated from the ectoderm by mesenchymal cells, no
Wnt-1 expression is observed (Fig. 8A). However, as the
underlying neuroepithelium approaches (Fig. 8B,C) and
eventually abuts the surface ectoderm (Fig. 8D), ectopic
activation of Wnt-1 expression is observed.
Our studies demonstrate that 3 of the 12 Wnt family
members examined are expressed in the primitive streak
and tailbud of the mouse embryo. Wnt-5a and Wnt-Sb
are expressed from the earliest stages of primitive streak
formation, whereas broad Wnt-3a expression is not de-
tected until the primitive streak is fully extended along
the proximodistal axis of the egg cylinder (7.5 dpc). As
well as temporal differences in the expression of these
family members, there are clear differences in the loca-
tion of Wnt transcripts in epiblast and mesodermal cells
within the primitive streak and tailbud. Thus, at egg
cylinder stages, Wnt-5a and Wnt-5b expression is pre-
dominant in the most posterior region of the primitive
streak; Wnt-5b is located medially in epiblast and me-
soderm, and Wnt-5a is medial-laterally located in me-
soderm. In contrast, Wnt-3a expression extends in me-
dial epiblast and mesoderm along the entire length of the
Cold Spring Harbor Laboratory Press on October 4, 2016 - Published by genesdev.cshlp.orgDownloaded from
Takada et al.
Figure 6. Whole-mount in situ hybridization of mesodermal markers to Wnt-Sa homozygous mutant embryos at 9.5 dpc.
+ / + or Wnt-3a
/ + embryos (identical results were obtained with wild-type or heterozygous embryos);
Mox-1 expression in somitic mesoderm. Hybridization is seen to all somites in
or Wnt-3a~/-^ embryos (arrow-
whereas Wnt-3a~/Wnt-3a~ embryos only show hybridization to somites rostral to the forelimb (FL).
Wnt-2 expression
in ventral mesoderm. Expression is normal, in Wnt-3a~/Wnt-3a' embryos, extending to the caudal limits of the truncated axis.
Biachyuiy expression in tailbud (large arrowhead) and notochord (small arrowhead). Normal notochord morphology terminates caudal
to the forelimbs in Wnt-3a~/Wnt-3a' embryos (large arrow). However, a few Brachyury-expTessing cells are visible in two discon-
tinuous streams on either side of the mutant embryo (small arrow in
Wnt-5a expression in the tailbud is entirely absent in
Wnt-3a ~/Wnt-3a
embryos (large arrowhead). In contrast, Wnt-5a is expressed normally in the CNS, branchial arches, and limbs. (HL)
primitive streak. In the late primitive streak and tailbud,
Wnt-3a and Wnt-Sb expression becomes localized to the
caudal-most region of the tail, whereas Wnt-5a expres-
sion extends rostrally, occupying a broad region at the
posterior end of the embryo. Assuming that Wnt tran-
scripts are translated at these stages and give rise to bi-
ologically active peptides, these results suggest that
there may be several diverse aspects to Wnt signaling in
the regulation of mammalian gastrulation.
Wnt expression domains and mesodermal fate
A clear picture of the normal fate of epiblast cells in
distinct regions of the primitive streak of the mouse em-
bryo at different stages has come from a variety of
ferent approaches, including single-cell labeling (Lav^son
et al. 1991; Lav^rson and Pedersen 1992), orthotopic graft-
ing (Beddington 1981, 1982, 1983; Tarn and Beddington
1992; Tam 1989), and culture of embryonic frag-
ments (Snow, 1981). These studies have established that
although cell types may not be irrevocably committed to
specific cell fates in different regions at the streak epi-
blast (Beddington 1981, 1982, 1983) there is a predictable
regional establishment of specific cell populations
within the gastrulating embryo.
A fate map for the 7.5-dpc late-streak mouse embryo
(R. Beddington, pers. comm.) is illustrated in Figure 9
relative to Wnt gene expression at the same stage. The
most posterior region of the primitive streak gives rise to
predominantly extraembryonic mesoderm, which lines
Cold Spring Harbor Laboratory Press on October 4, 2016 - Published by genesdev.cshlp.orgDownloaded from
Regulation of gastmlation
by Wnt-3a
Whole-mount in situ hybridization of CNS markers to
homozygous mutant embryos at 9.5 dpc.
{A,C,E,G] +
Wnt-3a''/+ embryos;
Expression of
is unaltered in the diencephalon of Wnt-
3a~/Wnt-3a~ embryos (arrowhead);
dorsal views of
expression in the somites (small arrowhead) and dorsal
spinal cord (large arrowhead).
expression shows normal dorsally restricted expression in the dorsal half of the CNS in Wnt-
3a~/Wnt-Sa" embryos except in the most caudal region where notochord development is disrupted.
{G,H) Wnt-1
expression at the
dorsal midline of the midbrain, hindbrain, and spinal cord is normal up to the level of the forelimb but becomes split laterally (small
arrow) in the kinked and folded neural tube of the
Wnt-3a ~/Wnt-3a "
embryo. Small isolated patches of neural tube-expressing
can also be seen (small arrowhead).
the yolk sac and gives rise to embryonic blood cells. In
contrast, the mid- and anterior streaks, at late egg cylin-
der stages, form exclusively embryonic mesoderm. The
mid-streak gives rise to the ventral (lateral) mesoderm
that lines body cavities and generates mesodermal deriv-
atives such as the kidney, whereas the anterior streak
forms the paraxial somitic mesoderm. The definitive en-
doderm and dorsal midline of the notochord and head
process are formed at the most extreme anterior end of
the streak from the node region.
Expression of Wnt genes in distinct regions of the
primitive streak correlates with regional diversity in the
formation of mesodermal subtypes in the mouse em-
suggesting that Wnt genes may regulate these pro-
cesses. Expression of Wnt-Sa and Wnt-5b, at egg cylinder
stages largely overlaps the region that gives rise to ex-
traembryonic mesoderm In contrast, Wnt-3a expression
extends into the region that generates embryonic meso-
derm. As there is some overlap in the expression do-
mains, there is no simple association of Wnt gene ex-
pression with extraembryonic and embryonic cell fates.
it is possible that some cells expressing Wnt-3a, or
within the Wnt-3a-expressing region, may give rise to
extraembryonic mesoderm. Similarly, some Wnt-5a/
Wnt-5t>-expressing cells may generate lateral mesoderm.
However, it is clear that the majority of cells expressing
Wnt-Sa/Wnt-Sb and Wnt-3a at egg cylinder stages do
have distinct fates in the conceptus.
At present, relatively little is known concerning the
biological properties of Wnt family members; however,
the few studies indicate that Wnt-5a and presumably
Wnt-5b, to which it shares much higher sequence con-
servation (87% amino acid identity) than to other Wnt
members (average 50% amino acid identity), are func-
tionally distinct from Wnt-3a. For example, Wnt-3a, like
will transform C57MG mammary epithelial
whereas Wnt-5a and Wnt-5b are inactive or only
weakly active (G. Wong, B. Gavin, and A. McMahon, in
Moreover, ectopic expression in Xenopus embryos
of Wnt-3a, but not Wnt-5a, results in axial duplication
(Moon et al. 1993; Wolda et al. 1993). Thus, it is tempt-
ing to suggest that Wnt-5a/Wnt-5b and Wnt-3a signaling
Cold Spring Harbor Laboratory Press on October 4, 2016 - Published by genesdev.cshlp.orgDownloaded from
Takada et al.
\ #1
Figure 8.
expression in transverse sections through the
caudal CNS of a Wnt-3a~/Wnt-3a~ mutant embryo.
open dorsal neural plate is visible in this and all other sections
and, as expected, expresses
at the unfused dorsal aspects
(small arrowhead). A more rostral region of closed neural tube
has folded under the open neural plate and expresses
(large arrow^head) where contact is made with the surface ecto-
No expression is seen elsewhere in the closed neural
More caudal sections through the same embryo. As
the neural tube moves toward
and eventually abuts
surface ectoderm, a new site of
expression occurs 180°
from the preexisting site of
expression (large arrowhead).
Expression is weak where the neural tube approaches the sur-
face ectoderm (5) but becomes strong when contact with the
surface ectoderm is established
(DA) Dorsal aorta; (G) gut.
during early gastrulation may be necessary for the gen-
eration of specific mesodermal subtypes. Interestingly,
Wnt-Sa and Wnt-Sb are all strongly expressed in the late
streak and tailbud stage embryo. At this stage there is no
new contribution of streak or tailbud precursors to ex-
traembryonic mesoderm (Tam and Beddington 1987).
Wnt-5a and Wnt-5b may have different roles dur-
ing early and late phases of gastrulation.
Regulation of embryonic mesoderm by Wnt-3a
Whereas the role of Wnt-5a and Wnt-5b remains a mat-
ter for speculation, Wnt-3a is clearly essential for em-
bryonic, but not extraembryonic, mesoderm formation.
A detailed analysis of mesodermal cell types in Wnt-3a
mutants at 9.5 dpc suggests that the earliest defect oc-
curs in the loss of somites at, and caudal to, the level of
the forelimb. Although the notochord is disrupted in
more posterior regions of the embryo, notochord devel-
opment at the forelimb level is normal. A subset of ven-
tral mesoderm that expresses Wnt-2 is largely unaffected
but terminates prematurely as a result of the axial trun-
cation that results from the loss of Wnt-3a expression.
These results suggest a number of possible functions for
Wnt-3a signaling during mouse gastrulation.
Wnt-3a expression extends, at late egg cylinder stages,
into anterior primitive streak cells fated to give rise to
the somites (Fig. 9). These appear to be the predominant
cell types expressing Wnt-3a alone. Thus, Wnt-3a may
be required to specify somitic cell fate. If this is the case,
then the most anterior seven to nine somites that form
normally are specified by a Wnt-3a independent path-
w^ay. Wnt-3a is not expressed at high levels until 7.5 dpc;
somites 1-7 may be specified prior to the activa-
tion of Wnt-3a. There is some support for this idea from
orthotopic grafting of labeled epiblast from the anterior
streak at 7.5 dpc (Tam and Beddington 1987). Grafted
cells only give rise to presomitic mesoderm after culture
to the somite stage six, suggesting that these labeled
cells will give rise to somites caudal to the sixth somite.
However, epiblast labeling by injection of wheat germ
agglutinin into the amniotic cavity at similar stages sug-
gests that labeled epiblast cells can contribute to somites
rostral to somite 6 (Tam and Beddington 1987). Addi-
tional grafting studies suggest that these cells may be
derived from more posterior-lateral regions adjacent to
the primitive streak (Tam 1989). The different outcomes
of these approaches presumably reflect technical differ-
ences in the experiments. Thus, at present it remains im-
clear whether somites that are independent of Wnt-3a
form prior to Wnt-3a expression. Interestingly,
Brachyury mutants that encode a defective T gene prod-
uct first show notochordal defects at a similar axial level
to Wnt-3a mutants (Beddington et al. 1992) even though
the Tgene is activated at the start of gastrulation. Thus,
there may be some common aspect to the developmental
program governing formation of different mesodermal
cell types that is initiated later in gastrulation.
From 7.5 dpc to 7.75 dpc, Wnt-3a expression moves
anterior in the primitive streak to terminate just poste-
rior to the node, the region that gives rise to the noto-
chord (Fig. 9; Tam and Beddington 1987; Lawson et al.
Thus, Wnt-3a, which encodes a secreted protein
Burrus and A. McMahon, in prep.), may directly reg-
ulate the development of notochord precursors in the
node, after initial notochord formation has commenced.
Interestingly, Wnt-3a has recently been shown to have
dorsalizing properties when ectopically expressed in Xe-
nopus embryos (Wolda et al. 1993) consistent with Wnt-
3a normally regulating notochord development. In Wnt-
Sa^ I'Wnt-3a~ embryos notochord morphogenesis is dis-
rupted. However, notochord cells are not entirely absent
but are present in low numbers and fail to form a midline
notochord, caudal to the forelimbs. Thus, there does not
appear to be an absolute requirement for Wnt-3a activity
to generate notochordal precursors at an axial level where
somites are entirely absent.
In addition to the somitic and notochordal disruptions,
Cold Spring Harbor Laboratory Press on October 4, 2016 - Published by genesdev.cshlp.orgDownloaded from
gastralation by Wnt-3a
Fate map Wnt expression
lateral mesoderm
head process / notochord / endoderm "
neural ectoderm
extraembryonic mesoderm
surface ectoderm
lateral mesoderm / neural ectoderm
lateral mesoderm / surface ectodenn
neural ectoderm / surface ectoderm
extraembryonic mesoderm / surface ectoderm
I Wnt-bh
I Wnt-ia I
Wnt-Sa Figure 9. Comparison of the fate map of
the late streak stage egg cylinder and Wnt
gene expression. The fate map is adapted
comm.) based on
data from Beddington
1982, 1983),
Tam and Beddington (1987), Tam (1989),
and Lawson et al.
no tailbud is formed in
homozygous mutant em-
The tailbud is established from cells deriving from
the mid-streak region (Snow 1981), a population of cells
localized midway along the proximodistal axis of Wnt-Sa
expression at 7.5 dpc. Thus, Wnt-Sa may play an addi-
tional role in establishing the mesodermal precursors
that become localized to the tailbud. Loss of this cell
population may then lead to axial truncation, a result
consistent with physical extirpation of the tailbud region
in the mouse embryo (Smith 1964).
Wnt-Sa may function in the assignment of several em-
bryonic mesodermal cell fates as discussed above. Alter-
natively, a similar phenotype may result if Wnt-3a is
required for normal mitogenic activity within the em-
bryonic mesodermal precursors of the primitive streak
and early tailbud. Cessation of cell proliferation would
be expected to lead to mesodermal and axial truncation
after all mesodermal precursors have differentiated. The
observed rostral-caudal loss of somites, then notochord,
and the apparently normal development of at least some
ventral mesodermal cell types would suggest that dis-
tinct regions of the embryonic mesoderm "run out" of
precursor cells before others. To distinguish between
these various hypotheses will require further work.
Moreover, it is clear that there are a large number of
other peptide signaling molecules and their receptors
that are expressed in the gastrulating mouse embryo, in-
cluding members of the FGF, TGF-p, and platelet-de-
rived growth factor families (for review, see Faust and
Magnuson 1993). Two of these have been shown to reg-
ulate mesoderm formation. Nodal mutants, which lack a
member of the TGF-p superfamily (Zhou et al. 1993),
show little evidence of mesoderm formation (Conlon et
1991; lannaccone et al. 1992). In contrast, mutants in
gene are adult viable, with disruption of only
the caudal-most mesodermal derivatives in the tail
(Mansour et al. 1993), despite the fact that
is ex-
pressed from 6.5 dpc in the primitive streak (Wilkinson
et. al. 1988). Thus, regulation of mesoderm formation is
likely to be complex, requiring the combinatorial action
of multiple factors operating at different developmental
stages. However, our studies establish a central role for
the Wnt-3a gene in regulation of embryonic mesoderm
formation and suggest that Wnt-5a and Wnt-Sb may
Cold Spring Harbor Laboratory Press on October 4, 2016 - Published by genesdev.cshlp.orgDownloaded from
Takada et al.
have distinct nonoverlapping functions during gastrula-
Wnt-3a and neural development
At 9.5 dpc, Wnt-3a is expressed along much of the dorsal
midline of the CNS and in a triangular "compartment"
in the diencephalon. Thus, Wnt-3a is implicated in di-
encephalic segmentation (Figdor and Stem 1993) as well
as the establishment of dorsoventral polarity in the CNS
(Parr et al. 1993). At and caudal to the forelimb, the neu-
ral tube is abnormally kinked and folded in Wnt-3a~ /
Wnt-Sa' embryos. Dysmorphology of the spinal cord
occurs precisely at the axial level where the somites are
absent and not in more rostral areas where Wnt-Sa is also
expressed. Thus, the somites may be required to physi-
cally constrain, or to regulate cell growth, in the neural
Interestingly, in Brachyuiy mutants, where somite
formation is disrupted, neural tube folding and kinking
are also observed. It appears that the neural tube contin-
ues to expand in Wnt-Sa mutants. In the absence of an
elongating trunk this may also lead to the observed fold-
ing and compression of the caudal CNS. Thus, it is likely
that the CNS dysmorphology occurs secondarily to the
loss of neighboring somites and truncation of the axis.
We have addressed the issue of dorsal patterning
within the Wnt-3a-deficient neural tube by examining
the expression of three dorsal markers,
and Pax-3 . All are expressed normally up to the point at
which CNS morphology is perturbed, suggesting that
Wnt-Sa is not required, at least to the forelimb level, for
CNS patterning. Severe disruption in CNS patterning is
observed in more caudal regions, although this is most
likely a secondary consequence of disruption of neural
tube morphogenesis and loss of the underlying noto-
In caudal regions, failure of the neural tube to close
leads to two dorsolateral domains, instead of a single
dorsal midline domain, of Wnt-1 and Pax-3 expression.
Loss of the notochord results in dorsal Pax-3 expression
moving ventrally as reported for notochordal removals in
the early chick embryo (Goulding et al. 1993). The most
interesting aspect of the pattern alteration is that of ec-
topic Wnt-1 expression in the neural tube. Loss of flank-
ing somites and compression of the neural tube allow
new points of contact to be made between the neural
tube and surface ectoderm, which lead to ectopic activa-
tion of
This result suggests that dorsal patterning
in the CNS may be initiated by ecotodermally derived
signals. As high-level expression of Wnt-1 occurs where
close apposition is seen between the CNS and surface
ectoderm, these signals may act quite locally, in an anal-
ogous fashion to those of the notochord, which induce
the ventral floor plate (Jessel and Melton 1992).
In summary, the CNS phenotypes are best explained
as secondary consequences of the loss of mesodermal
derivatives rather than the loss of dorsal Wnt-3a expres-
sion in the CNS. Wnt-3a and Wnt-1 colocalize in much
of the dorsal CNS, and we have suggested that they may
have a redundant function (McMahon