![[Feedback]](/icons/banner/feedbackACT.gif){kind=icon}
![[For Subscribers]](/icons/banner/subscriberACT.gif){kind=icon}
![[Archive]](/icons/banner/archiveACT.gif){kind=icon}
![[Search]](/icons/banner/searchACT.gif){kind=icon}
![[Contents]](/icons/banner/tocACT.gif){kind=icon}
|
|
|
|
|
||
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Vol. 12, No. 11, pp. 1571-1586, June 1, 1998
Department of Anatomy and Program in Developmental Biology, School of Medicine, University of California at San Francisco, San Francisco, California 94143-0452 USA
"Fibroblast growth factor" (FGF) was first
identified 25 years ago as a mitogenic activity in pituitary extracts
(Armelin 1973 The purpose of this review is to discuss the functions performed by
members of the FGF family in one of the best-studied vertebrate developmental systems Initial efforts to purify the factors in brain tissue that
stimulated fibroblast proliferation led to the identification of two
closely related proteins
Introduction
Top
Introduction
References
; Gospodarowicz 1974). This modest observation
subsequently led to the identification of a large family of proteins
that affect cell proliferation, differentiation, survival, and motility
(for review, see Basilico and Moscatelli 1992; Baird 1994). Recently, evidence has been accumulating that specific members of the FGF family
function as key intercellular signaling molecules in embryogenesis (for
review, see Goldfarb 1996). Indeed, it may be no exaggeration to say
that, in conjunction with the members of a small number of other
signaling molecule families [including WNT (Parr and McMahon 1994),
Hedgehog (HH) (Hammerschmidt et al. 1997), and bone morphogenetic
protein (BMP) (Hogan 1996)], FGFs are responsible for inducing
and/or regulating the subsequent development of most organs in the vertebrate body.
limb formation. It focuses initially on what is
known about FGF function in the established limb bud and then discusses
the mechanisms by which the signaling centers that control outgrowth
and patterning of the established limb bud are formed, emphasizing
possible roles played by FGFs in these processes. Finally, it discusses
the potential role of FGFs in the induction of limb development. The
insights into the requirement for FGF signaling in bone development
that have been gained from analyses of mutations in human and mouse FGF
receptor (FGFR) genes will not be discussed, as those studies have been
the subject of several recent reviews (Wilkie et al. 1995; Yamaguchi
and Rossant 1995; De Moerlooze and Dickson 1997; Webster and Donoghue 1997).
FGF ligand and receptor families
acidic FGF and basic FGF (now designated FGF1
and FGF2, respectively). With the advent of gene isolation techniques
it became apparent that the Fgf1 and Fgf2 genes are members of a large family, now known to be comprised of at least 17 genes, Fgf1-Fgf17, in mammals (see Coulier et al. 1997;
McWhirter et al. 1997; Hoshikawa et al. 1998; Miyake 1998). At least
five of these genes are expressed in the developing limb (see Table 1). The proteins encoded by the 17 different FGF
genes range from 155 to 268 amino acid residues in length, and each
contains a conserved "core" sequence of ~120 amino acids that
confers a common tertiary structure and the ability to bind heparin or
heparan sulfate proteoglycans (HSPGs) (Zhu et al. 1991; Faham et al.
1996). Although all family members share significant amino acid
sequence identity in the core region, the sequences flanking it are
generally poorly conserved. The orthologs of many of the mammalian FGF
genes have been identified in other vertebrate species, and recently, FGF family members have been identified in worms (Burdine et al. 1997,
1998) and flies (Sutherland et al. 1996). Consistent with their
potential functions as intercellular signaling molecules, many of the
FGFs are exported efficiently from the cells that produce them. Once
released from cells, FGFs bind avidly to HSPGs such as the syndecans,
glypican, and perlecan on the cell surface and in the extracellular
matrix (ECM), which is thought to restrict their ability to diffuse
very far from the cells that produced them (for review, see Basilico
and Moscatelli 1992).
Table 1.
Expression domains of FGF and FGFR genes in the
developing limb
The binding of FGFs to HSPGs (sometimes termed low-affinity FGFRs)
facilitates FGF signal transduction by oligomerizing and presenting the
ligands to high-affinity FGFRs (Faham et al. 1996; for review, see
Mason 1994), which are transmembrane protein tyrosine kinases. In
vertebrate species, a family of four genes, Fgfr1-Fgfr4, encodes such high-affinity FGFRs, each of which is capable of producing
a variety of receptor RNA isoforms through alternative splicing (for
review, see Johnson and Williams 1993). Two of these genes apparently
play a role in early limb development (Deng et al. 1997; Xu et al.
1998; see Table 1). The binding of FGFs activates the high-affinity
receptor proteins by inducing the formation of receptor homo- or
heterodimers, which results in receptor transphosphorylation and leads
to the activation of a RAS-dependent intracellular signal transduction
pathway (for review, see Fantl et al. 1996).FGFRs can also be
activated by ligands other than FGFs (for review, see Green et al.
1996), but the biological significance of those findings is not known.
Also, it remains to be determined how the activation by different FGFRs
of a common signal transduction pathway results in a multiplicity of
cellular responses.
The specificity of FGF binding to FGFRs is determined by sequences in
the extracellular domain of the receptor proteins, particularly in a
region close to the transmembrane domain containing an
immunoglobulin-like loop (termed Ig-loop III) that can be varied by
alternative splicing (for review, see Johnson and Williams 1993). When
assayed in vitro, individual FGFR proteins bind multiple FGFs but also
display a unique pattern of affinities for the different ligands (for
review, see De Moerlooze and Dickson 1997). However, as ligand binding is also greatly influenced by the distribution of HSPGs at the cell
surface and in the ECM, it is unknown to what extent these in vitro
assays reflect the ligand binding specificity of different FGFR
proteins in vivo.
| |
Signaling centers in the established limb bud |
|---|
Much of what is known about the early stages of vertebrate limb development has been learned from experimental studies of the chick embryo, which is manipulated in ovo easily. However, genetic studies in the mouse have provided important insights into the roles played by specific genes in limb development, and comparative studies of gene expression and function during chick and mouse limb development have served to demonstrate that the basic mechanisms of limb formation have been evolutionarily conserved. The description of limb development that follows is a composite of information from studies of both chick and mouse embryos.
Limb development is heralded by the protrusion from the lateral body
wall of a small "bud," which is comprised of lateral plate
mesoderm (LPM) cells and the overlying surface ectoderm (Fig.
1A). The mesenchymal cells in this bud proliferate
and eventually give rise to the skeletal elements and other connective
tissue of the mature limb, whereas the limb muscles are derived from cells that migrate into the limb bud from the somites (Chevallier et
al. 1977; Christ et al. 1977). As development proceeds the limb
elongates along its proximal-distal (P-D) axis (shoulder to fingers),
becomes flattened along its dorsal-ventral (D-V) axis (back of hand
to palm), and asymmetric along its anterior-posterior (A-P) axis
(thumb to little finger). Differentiation becomes morphologically apparent as the mesenchymal cells condense to form the primordia of
individual skeletal elements. The most proximal element, designated the
stylopod, begins to differentiate first, with its proximal end forming
before the distal end, followed at successively later times by the
progressive differentiation of more distal structures (zeugopod and
autopod; see Fig. 2A).
|
|
Such outgrowth and patterning depends on the establishment and
maintenance of three discrete signaling centers within the limb bud:
(1) the apical ectodermal ridge (AER or ridge), a morphologically distinct epithelium that runs from anterior to posterior at the distal
margin of the bud (Fig. 1B); (2) the zone of polarizing activity (ZPA
or polarizing region) in the mesenchyme at the posterior margin of the
bud, which has no distinguishing morphological features; and (3) the
nonridge ectoderm of the bud (Fig. 1C). The functions of the signals
from these regions have been determined by experimental studies in the
chick. Elimination of the AER results in a failure of subsequent P-D
outgrowth (Saunders 1948; Summerbell 1974) (see Fig 2B), indicating
that signals from the ridge are continuously required for outgrowth of
the limb. However, these signals per se do not provide the P-D
patterning information that functions to shape the limb skeletal
elements (Rubin and Saunders 1972). Grafting the ZPA to the anterior
side of a host limb bud results in mirror-image duplication of the
posterior limb, indicating that signals from the ZPA regulate
patterning along the limb A-P axis (Saunders and Gasseling 1968).
Rotation of limb bud ectoderm through 180° results in a reversal of
only D-V polarity, at least of the distal part of the experimental
limb (MacCabe et al. 1974; Pautou 1977), indicating that signals from
the nonridge ectoderm regulate patterning along the limb D-V axis.
Some of the secreted molecules responsible for the activities of the
different limb bud signaling centers have been identified. As discussed
below, FGFs produced by AER cells perform the functions of the ridge
that are required for P-D outgrowth. Sonic hedgehog (SHH) produced by
ZPA cells is the key mediator of the polarizing activity that regulates
patterning along the A-P axis (Riddle et al. 1993; Chang et al. 1994).
WNT7A, produced by dorsal ectoderm cells and acting through its
downstream target gene Lmx1 expressed in the underlying
mesenchyme, plays a role in patterning along the D-V axis (Parr and
McMahon 1995; Riddle et al. 1995; Vogel et al. 1995). It is possible
that the ventral ectoderm also secretes a signaling molecule involved
in D-V patterning, but as yet no candidate signal has been identified.
Importantly, these signals are interdependent. There are regulatory
interactions among the different signaling centers, and their products
work cooperatively to regulate and coordinate limb outgrowth and
patterning along all three axes. For example, as discussed below, FGFs
from the ridge and WNT7A from the dorsal ectoderm are required to
maintain Shh expression (and, hence, ZPA activity), and SHH in
turn influences FGF gene expression in the ridge.
These signals act on a population of undifferentiated mesenchymal cells
located in a region known as the "progress zone," situated near
the distal tip of the limb bud, subadjacent to the AER (see Fig. 1C).
As the progress zone cell population proliferates and the limb
elongates, some of the cells are left behind (proximal to) the progress
zone and are therefore no longer exposed to the signals produced by the
AER. To account for the mechanism by which cells are patterned along
the P-D axis, Summerbell et al. (1973) proposed that time of exit from
the progress zone determines cell fate. Time spent in the progress zone
might be measured in terms of the number of divisions a cell has
undergone, or perhaps in relation to a molecular clock based on the
total amount of AER-derived signal to which it has been exposed. While
they are in the progress zone, cells are also receiving the signals
that confer A-P and D-V positional information, which help to
determine precisely what structures they ultimately form. Despite the
advances in identifying the molecules produced by the different limb
bud signaling centers, the molecular mechanism by which progress zone
cells interpret such signals is unknown.
| |
FGF function in the established limb bud |
|---|
As noted above, the AER is essential for limb outgrowth, and its
position and size are critical determinants of limb form. For example,
when the AER is rotated 90°, outgrowth takes place at right angles
to the original axis and conforms to the new position of the AER
(Zwilling 1956). Moreover, the dimensions of the AER seem to determine
the number of digits that develop, as the formation of supernumerary
digits along the A-P axis is always correlated with an increase in the
A-P length of the ridge (see, e.g., Goetinck and Abbott 1964; MacCabe
et al. 1975; Lee and Tickle 1985). Of the 17 known FGF genes, 5 are
expressed in the distal part of the established limb bud (see Table 1),
4 in the AER (Fgf2, Fgf4, Fgf8, Fgf9; e.g., see Fig. 1C), and
2 in the mesenchyme underlying it (Fgf2, Fgf10), and their
role is apparently to provide AER function, either directly or indirectly.
FGFs perform the functions of the AER
The many genes expressed in the AER include members of several
different families that encode secreted signaling molecules. Given the
potential complexity of AER function and the variety of possible
candidate molecules that might perform its functions, it came as
somewhat of a surprise when it was first discovered that an individual
FGF protein could substitute for the AER. Thus, when heparin beads
soaked in recombinant FGF protein (FGF beads) were applied to the
exposed limb bud mesenchyme following AER removal at early stages of
chick limb development in ovo, the limbs that subsequently formed were
essentially normal (Niswander et al. 1993; Fallon et al. 1994). For
example, when two FGF4 beads were applied simultaneously, ~80% of
the limbs developed a full complement of skeletal elements including
the autopod, whereas most control limbs failed to form both the
zeugopod and autopod (Niswander et al. 1993; see Fig. 2C).
The observation that the proteins encoded by at least three of the FGF
genes expressed in the AER, FGF2 (Fallon et al. 1994), FGF4 (Niswander
et al. 1993), and FGF8 (Vogel et al. 1996), can each substitute for the
AER in ovo, raises the possibility that all the FGFs produced in the
AER contribute to AER function, perhaps by each providing a fraction of
the total FGF required. Thus, application of any one FGF protein might
rescue development of a ridgeless limb when it is supplied in a
quantity equal to or greater than the total amount of all FGFs normally
produced in the AER. Although all FGF genes expressed in the AER may
contribute to its function, recent studies have shown that embryos
homozygous for null alleles of either Fgf2 (Ortega et al.
1998; Zhou et al. 1998; R. Dono, G. Texido, R. Dussel, H. Ehmke, and R. Zeller, in prep.) or Fgf9 (D. Ornitz, pers. comm.) have normal
limbs. These results indicate that neither of these genes is required for normal limb formation but do not exclude the possibility that they
perform functions that are redundant with those of other FGF genes
expressed in the limb bud.
It has not been possible to demonstrate a requirement for either
Fgf4 or Fgf8 in limb development by standard genetic
methods, because embryos homozygous for null mutations in either of
these genes die before the limbs are formed (Feldman et al. 1995;
Meyers et al. 1998). However, the problem of the early embryonic
lethality of Fgf8 mutant embryos has been circumvented in
studies employing the Cre/loxP binary system
(for review, see Kilby et al. 1993; see also Gu et al. 1994) to obtain
tissue-specific inactivation of Fgf8. By mating mice carrying
an Fgf8 allele in which vital exons of the gene are
"floxed" (flanked by loxP
sites, the recognition sites for Cre recombinase) (Meyers et al. 1998)
to mice carrying a cre gene expressed specifically in the AER,
it has been possible to obtain embryos in which all functional
Fgf8 gene expression has been eliminated from the AER by
embryonic day (E)10.5 (~35 somites) (M. Lewandoski, E. Meyers, Y.-H.
Liu, R. Maxson, and G. Martin, unpubl.). The mutant animals display
consistent skeletal abnormalities in both forelimbs and hindlimbs,
including loss of digits. The data provide evidence that Fgf8
gene function in the AER is necessary for normal limb development but
suggest that other FGF genes are also required. Further studies,
including AER-specific inactivation of Fgf4, should help to
elucidate the requirements for FGF gene function in the ridge. It will
also be important to determine what functions are performed by other (non-FGF) genes in the AER.
Specific functions of FGFs produced in the AER
There is substantial evidence that FGFs produced in the AER serve
at least two major functions. One function is to stimulate the
proliferation of cells in the progress zone and thus produce the new
cells required for limb outgrowth. Studies employing markers for cell
proliferation have directly demonstrated that FGFs are mitogens for
limb bud mesenchyme (Aono and Ide 1988; Niswander and Martin 1993;
Niswander et al. 1993; Dealy et al. 1997). Moreover, ectopic expression
of Fgf2 on the anterior side of the wing bud apparently causes
excess proliferation of mesenchymal cells, resulting in abnormal
thickening and splitting of the skeletal primordia (Riley et al. 1993).
Although in vitro studies have suggested that the ability of FGFs to
promote the outgrowth of limb mesenchyme is dependent on IGF-I (Dealy
et al. 1996), limb development is not specifically affected in mice
that lack functional IgfI and IgfII genes (Liu et al.
1993), thus indicating that IGF signaling is not required for any
aspect of normal limb development. There is also evidence that FGFs are
required, in conjunction with SHH protein, to activate the expression
of Hox genes and Bmp2 in the mesenchyme (Laufer et
al. 1994; Niswander et al. 1994), but this function may be related to
their effects on cell proliferation (Duboule 1995; Ohsugi et al. 1997).
Another major function of FGFs produced in the AER is to maintain
Shh expression in the ZPA. Initial studies of this FGF
function relied on assays for polarizing activity that involved
transplanting cells from the posterior side of an experimental limb bud
to the anterior side of a host limb bud and assessing their ability to induce posterior skeletal duplications. This type of analysis demonstrated that polarizing activity was lost following AER removal but that it was maintained when an FGF bead was applied to the ridgeless posterior limb bud mesenchyme (Niswander et al. 1993; Vogel
and Tickle 1993). After SHH was identified as the key mediator of
polarizing activity, several FGFs were shown to be capable of
substituting for the AER to maintain Shh expression (Laufer et
al. 1994; Niswander et al. 1994; Crossley et al. 1996; Vogel et al.
1996; Ohuchi et al. 1997) (Fig. 2D,E).
Such experiments leave open the question of which FGFs maintain the ZPA
in the normal limb bud. The available evidence indicates that although
FGF4 is not required to induce the expression of Shh, it is
largely, but perhaps not exclusively, responsible for maintaining
Shh expression as the limb elongates. Thus, in the limb buds
of mouse embryos homozygous for mutant alleles of the limb
deformity (ld) gene, which fail to express Fgf4,
Shh RNA is detected at early stages of limb bud development (up to
E10.5) but, subsequently, is detected at very low levels or not at all (Chan et al. 1995; Haramis et al. 1995; Kuhlman and Niswander 1997).
However, because the level of Fgf8 expression is reduced in
the ld mutant AER, the loss of Shh expression may not
be due solely to the absence of FGF4. It has yet to be determined
whether Shh levels are reduced following loss of Fgf8
function in the AER.
The dependence of Shh expression on FGFs produced in the AER
provides an explanation for the observation that as the limb elongates,
the ZPA is always located near the ridge. However, it is not known
whether the ZPA is comprised of a stable cell population, which always
remains distal because cells exiting from the progress zone move
proximal to it, or whether it is a transient cell population, which is
continuously recruited from the posterior progress zone and ceases to
express Shh as it is displaced proximally. The observation
that cells proximal to the ZPA can be induced to express Shh
in response to an FGF bead has been interpreted as reflecting a
reactivation of Shh gene expression in cells that had left the
ZPA (Yang and Niswander 1995) and thus supporting the hypothesis that
the ZPA is a transient cell population. However, conclusive evidence on
this point will require a cell lineage analysis.
It is possible that the regulatory interactions between Fgf4
and Shh are reciprocal, as ectopic expression of Shh
in the anterior mesenchyme leads to ectopic expression of Fgf4
expression in the overlying anterior AER. Thus, it has been proposed
that in the normal limb bud, SHH produced in the ZPA initially induces
and subsequently maintains Fgf4 expression in the AER (Laufer
et al. 1994; Niswander et al. 1994), perhaps via Bmp2 (Duprez
et al. 1996). Consistent with this hypothesis, it has been found that although an AER forms in the limb buds of mouse embryos homozygous for
a null allele of Shh, it does not express Fgf4
(Chiang et al. 1996; C. Chiang, pers. comm.). Interestingly,
Fgf8 is expressed in the AER of Shh mutant limb buds,
supporting the hypothesis that Fgf8 functions upstream of
Shh (see below). A positive feedback loop between
Fgf4 and Shh provides one mechanism by which
outgrowth and patterning of the limb could be coordinately regulated,
but it is important to bear in mind that other secreted signaling molecules, such as WNT7A produced by the dorsal ectoderm, may also play
a role in regulating Shh expression (Parr and McMahon 1995;
Yang and Niswander 1995).
In addition to Shh, there are a large number of genestoo
many to discuss herewhose expression in the limb bud mesenchyme is
maintained by FGF signaling from the AER. Several of these are
transcription factors that may play an important role in interpreting the early patterning signals and initiating the cascade of gene expression that ultimately results in the development of a limb (for
review, see Tickle and Eichele 1994). One of the challenges for future
studies is to determine which of these genes is directly regulated by FGFs
from the AER and how they function to maintain limb outgrowth and patterning.
Function of FGF10 produced in the limb bud mesenchyme
Fgf10 is one of the genes expressed in the distal limb
bud mesenchyme that may be a direct target of FGF signaling from the AER (Ohuchi et al. 1997). As discussed below, Fgf10 is thought to play a key role at very early stages of limb development, prior to
the development of the AER. In the established limb bud, Fgf10 expression, which is normally detected at high levels in the distal mesenchyme, is no longer observed 10 hr after ridge removal. Implants of Fgf8-expressing cells can rescue Fgf10 expression.
On the basis of these observations, as well as the finding that
Fgf10-expressing cells can induce both Fgf8
expression and thickening of nonridge ectoderm, Ohuchi et al. (1997)
have proposed that there may be a positive feedback loop between
Fgf8 and Fgf10, similar to the one proposed for
Fgf4 and Shh. Moreover, they have suggested that FGF10 might be the factor produced by the distal mesenchyme that serves
to maintain the AER, that is, the "AER maintenance factor," the
existence of which has been inferred from classic embryological studies
in the chick embryo (for review, see Hinchliffe and Johnson 1980).
| |
Formation of the ZPA and AER |
|---|
Knowledge of the molecules and cellular mechanisms involved in forming the signaling centers present in the established limb bud is crucial to understanding limb development. In the chick, ZPA formation, as reported by Shh expression, begins at approximately the same stage as outgrowth of the limb bud first becomes evident. A few hours later, morphogenetic changes occur in the surface ectoderm, resulting in the formation of the AER, which contains cells that are both morphologically and molecularly distinct from nonridge ectoderm cells. Several lines of evidence suggest that FGF genes, particularly Fgf8 and Fgf10, play a role in mediating the initial outgrowth of the limb and in establishing the ZPA and AER.
Initiation of limb bud outgrowth
The bulge in the lateral wall of the embryo that marks the onset
of limb bud formation results from the continual proliferation of cells
in the prospective limb-forming region and a concomitant decrease in
cell proliferation elsewhere along the length of the LPM (Searls and
Janners 1971). Because FGF8 can stimulate outgrowth of the LPM (Mahmood
et al. 1995a; Crossley et al. 1996; Vogel et al. 1996) and is expressed
before limb bud outgrowth is evident, it has been proposed that FGF8
produced in the surface ectoderm functions to maintain the
proliferation of the LPM in the prospective limb-forming regions. This
hypothesis is supported by the observation that there is no expression
of Fgf8 in the prospective limb ectoderm and no initiation of
limb bud outgrowth in mouse embryos homozygous for a mutant allele of
Fgfr2 (Xu et al. 1998). However, in limbless mutant
chick embryos there is apparently normal initial outgrowth of the limb
bud in the absence of Fgf8 expression in the ectoderm (Grieshammer et al. 1996; Noramly et al. 1996; Ros et al. 1996). This
observation provides evidence that FGF8 is not required, at least in
the limbless mutant, for the initial phase of limb bud
outgrowth but does not exclude the possibility that FGF8 contributes mitogenic activity in the normal limb bud.
Based on its expression pattern, the Fgf10 gene is a candidate
for stimulating early limb bud outgrowth. At early stages of development Fgf10 expression is detected widely in the LPM but then is down-regulated in the interlimb region and becomes restricted to the mesoderm of the prospective forelimb and hindlimb territories just prior to the start of limb bud outgrowth (Ohuchi et al. 1997). However, there is evidence suggesting that FGF10 acts through an FGFR2
isoform (IIIb) that is expressed, at least at later stages of
development, by epithelial rather than mesenchymal cells (see Ohuchi et
al. 1997; Xu et al. 1998). Further studies are required, including
inactivation of Fgf10 in the prospective limb bud mesoderm and
Fgf8 in the overlying ectoderm, to directly determine the roles of
these and other genes in regulating the initial outgrowth of the limb bud.
Formation of the ZPA
An important issue in early limb development is how the ZPA is
established and localized to the posterior limb bud (for review, see
Johnson and Tabin 1997). Just prior to the stage at which limb bud
outgrowth begins in the chick embryo, polarizing activity (competence
to express Shh and thus form a ZPA) is detected not only in
prospective posterior limb territory but also in the interlimb region
(Hornbruch and Wolpert 1991; Stratford et al. 1997). It appears that
the clustered Hox genes, such as Hoxb8, have a role in providing such competence to express Shh (Charité et
al. 1994; Lu et al. 1997; Stratford et al. 1997). However, there is
substantial evidence that in the competent regions FGF signaling is
necessary to induce Shh expression, and that during normal
limb development FGf8 is the specific family member that serves as the
required FGF.
This conclusion is based on several findings. Fgf8 is
expressed prior to Shh expression during the development of
both normal and ectopic limb buds (Crossley et al. 1996; Vogel et al.
1996). In either Fgfr2
IgIII (Xu et
al. 1998) or limbless (Grieshammer et al. 1996; Noramly et al.
1996; Ros et al. 1996) mutant embryos, which fail to express Fgf8, no Shh is detected. Moreover, FGF8 can induce
Shh expression in limbless limb buds (Grieshammer et
al. 1996). Taken together, these data indicate that FGF8 produced in
the surface ectoderm has a role in inducing Shh expression in
the underlying mesoderm and that restriction of Fgf8
expression to the prospective limb ectoderm serves to restrict ZPA
formation to the prospective limb mesenchyme. The posterior restriction
of Shh expression within the nascent limb bud is due either to
a lack of competence in the anterior limb bud mesenchyme to respond to
FGF8 produced in the overlying ectoderm or to the presence of a
repressor of Shh expression (for review, see Johnson and Tabin
1997; see also Qu et al. 1997).
Localization of the prospective AER
The data discussed above emphasize the importance for normal limb development of appropriately restricting Fgf8 expression specifically to the ectoderm of the prospective limb territory. It has long been assumed that the AER, for which Fgf8 is an early molecular marker, forms at the boundary between dorsal and ventral territories, but until relatively recently there was no means of confirming this assumption or exploring the molecular mechanism by which the prospective site of AER formation is determined. This problem has been solved to some extent by the identification of markers for D-V polarity in the developing limb bud.
Wnt7a (Dealy et al. 1993; Parr et al. 1993) and En1
(Davis et al. 1991), a vertebrate homolog of the Drosophila
transcription factor gene engrailed, are expressed in the
dorsal and ventral ectoderm of the established limb bud, respectively.
Expression of both genes is initially detected prior to formation of
the AER, and both are required for normal D-V patterning of the distal limb (Parr and McMahon 1995; Loomis et al. 1996; Cygan et al. 1997;
Loomis et al. 1998). Using these two genes as markers, it has not only
been confirmed that the AER is located at the D-V boundary of the limb
bud, but it has also been shown that the AER itself is divided into a
dorsal domain in which neither Wnt7a nor En1 are
expressed, and a ventral one in which En1 is expressed (Fig.
3A,B). As yet, genes specifically expressed in the
dorsal AER have not been identified.
|
Wnt7a and En1 have also been useful as markers for
demonstrating that the mechanism by which Fgf8 expression is
induced in the prospective limb ectoderm is in some way linked to the
establishment of normal D-V polarity. Thus, it has been found that in
embryos homozygous for the chick limbless mutation there is
both a failure to express Fgf8 in the limb bud ectoderm and a
lack of normal D-V polarity (Grieshammer et al. 1996; Noramly et al.
1996; Ros et al. 1996). A relationship between AER formation and D-V
polarity is further demonstrated by the observations that
transplantation of prospective dorsal tissue into ventral limb domains,
or vice versa, results in the formation of an ectopic AER (Tanaka et
al. 1997), and replacing patches of dorsal ectoderm with ventral
ectoderm leads to ectopic expression of Fgf8, presumably at
the border between host dorsal and graft ventral tissue (Laufer et al. 1997).
Ectopic gene expression studies in the chick have suggested that it is
not apposition of dorsal and ventral territories per se that is
required to induce AER formation, as suggested previously by Meinhardt
(1983), but instead that the AER forms as a consequence of a cascade of
gene expression that is induced at the boundary between cells in the
dorsal ectoderm that express Radical-fringe (R-fng),
one of three vertebrate homologs of the Drosophila fringe gene
(Wu et al. 1996; Cohen et al. 1997; Johnston et al. 1997; Sakamoto et
al. 1997), and nonexpressing cells in the ventral ectoderm (Laufer et
al. 1997; Rodriguez-Esteban et al. 1997). This hypothesis is appealing
because it suggests that the function of the fringe gene,
which plays a key role in the formation of the wing margin required for
outgrowth and patterning of the Drosophila wing (for review,
see Brook et al. 1997; Irvine and Vogt 1997), has been evolutionarily
conserved. By further analogy with Drosophila, it has been
suggested that R-fng may function in AER formation via effects
on the Notch signaling pathway (Laufer et al. 1997; Rodriguez-Esteban
et al. 1997). Although the model based on R-fng gain-of-function experiments in the chick is attractive, data from
other types of experiments are not consistent with the hypothesis as
proposed (for review, see Irvine and Vogt 1997). In the absence of
loss-of-function data, a requirement for R-fng function in AER
formation must remain speculative.
Mechanism of AER morphogenesis
At prelimb bud stages, fate-mapping studies show that AER
precursors are widely distributed in prospective limb ectoderm (Altabef et al. 1997; Michaud et al. 1997). However, by the stage at which morphogenesis of the ridge begins (stage 18 in the chick), most progenitors of the AER appear to be localized primarily along the
distal and ventral surface of the nascent limb bud (Michaud et al.
1997) (Fig. 3C,D). Gene expression studies in both chick and mouse
embryos show that just prior to formation of a morphologically distinct
ridge, cells expressing molecular markers of the AER such as Fgf8,
Dlx2, and CD44, are similarly localized (Crossley and Martin 1995;
Crossley et al. 1996; Loomis et al. 1998; S. Bell, C.M. Schreiner, and
W.J. Scott, in prep.). The first visible manifestation of ridge
morphogenesis in the mouse limb bud is a change in the shape (from flat
to cuboidal) of the presumptive AER progenitors throughout the
distal-ventral ectoderm. Subsequently, it appears that they continue
to undergo changes in cell shape and to move progressively more
distally, so that they ultimately become concentrated in a narrow ridge
at the distal margin of the limb bud, which consists of a layer of
pseudostratified columnar basal cells and several layers of superficial
flattened cells (Milaire 1974; Loomis et al. 1998; S. Bell, C.M.
Schreiner, and W.J. Scott, in prep.). In the chick, these morphological
changes appear to occur more rapidly than in the mouse, and the AER
that forms consists of a pseudostratified basal layer and a single, flattened superficial layerthe periderm (Todt and Fallon 1984).
Loss of En1 function inhibits the process by which the
cuboidal AER precursors in the ventral ectoderm change shape and become localized in a narrow ridge (Loomis et al. 1996). On the basis of a
detailed analysis of En1 mutant embryos, Loomis et al. (1998) concluded that the cells that form the dorsal domain of the AER (i.e.,
those that do not express En1) are not inhibited from
undergoing most of the normal events of AER morphogenesis; therefore, a
normal dorsal AER forms in the En1 mutant limb bud, whereas
the ventral AER progenitor cells remain spread throughout the ventral
ectoderm. Interestingly, the En1 mutant limbs display largely
normal P-D and A-P outgrowth and patterning, suggesting that
prospective ridge cells do not need to form a morphologically normal
ridge to perform their functions. On the basis of these observations, it was proposed that the morphogenesis of the normal AER depends on
behavioral differences between the progenitors of the dorsal and
ventral domains of the AER, and that dorsal prospective AER cells may
stimulate ventral prospective AER cells to move towards them.
An analysis of En1/Wnt7a double mutant embryos
(Cygan et al. 1997; Loomis et al. 1998) has led to the suggestion that
the proposed dorsal behavior of prospective AER cells is specified by
some form of WNT signaling but that Wnt7a is unlikely to be the WNT family member that provides the presumed dorsalizing signal in
the normal limb bud. It is also possible that signaling via the Notch
receptor is in some way involved in specifying the proposed differences
between dorsal and ventral prospective AER cells, as the mouse mutation
syndactylism, which causes abnormalities in AER morphogenesis,
appears to be a hypomorphic allele of Serrate2, a homolog of
the Drosophila gene that encodes the Notch ligand Serrate
(Sidow et al. 1997).
Fgf8 expression in prospective AER cells may also have an
important role in ridge morphogenesis. Fgf8 RNA is detected in
prospective AER cells either just prior to or shortly after they begin
to thicken (Crossley and Martin 1995; Crossley et al. 1996; Loomis et
al. 1998; S. Bell, C.M. Schreiner, and W.J. Scott, in prep.), although
it is not known whether Fgf8 is expressed in all cells that
will ultimately form the ridge. On the basis of the recent finding from
studies of Fgf8 null mutant mouse embryos that Fgf8 is required during gastrulation to enable nascent mesoderm cells to
move away from the primitive streak (X. Sun, E. Meyers, M. Lewandoski,
and G. Martin, unpubl.), it is tempting to speculate that FGF8 produced
in prospective AER cells mediates or facilitates the morphogenetic
movements that occur during ridge formation. This cannot be determined
from studies of embryos in which Fgf8 has been inactivated
after the AER has formed (M. Lewandoski, E. Meyers, Y.-H. Liu, R. Maxson, and G. Martin, unpubl.) but could be studied if Fgf8
were inactivated in the prospective limb bud ectoderm prior to AER formation.
Any molecular model of AER formation should include an "AER
inducer," whose presence in the mesoderm of the prospective limb region has been inferred from experimental studies in the chick. Both
this inducer in the mesoderm and competence of the overlying ectodermal
cells to respond to it are present only transiently (for review, see
Hinchliffe and Johnson 1980; see also Carrington and Fallon 1984).
FGF10 appears to have some activities consistent with a role as the
mesoderm-derived inducer (Ohuchi et al. 1997). Presumably it alone does
not perform this function, as AER-inducing activity is not detectable
in the established limb bud, whereas Fgf10 continues to be
expressed at high levels in the distal limb bud mesenchyme until late
stages of limb development. Alternatively, it is possible that FGF10 is
itself the mesoderm-derived AER inducer but that an inhibitor of its
activity is present at later stages of limb development.
| |
FGF function in limb bud induction |
|---|
An important question about limb development is what activates the
process and induces the limbs to form where they do. As with other
classic examples of embryonic induction, both an inducer and competence
to respond to the inducer are necessary to initiate limb development.
Remarkably, it has been found that an individual FGF protein is
sufficient to induce the formation of a complete limb (Fig.
4A,B). Thus, Cohn et al. (1995) showed that insertion of an FGF bead in the LPM of the interlimb region (often referred to as
"the flank") of a chick embryo in ovo results in the formation of
an ectopic limb, provided that the bead is applied prior to the stage
at which normal limb bud outgrowth begins (stage 17).
|
FGF-induced ectopic limbs
A number of FGF family members can induce an ectopic limb,
including FGF1, FGF2, FGF4, FGF8, and FGF10 (Cohn et al. 1995; Ohuchi
et al. 1995; Crossley et al. 1996; Vogel et al. 1996; Ohuchi et al.
1997), although differences have been observed with respect to the
efficiency and the stage specificity of their effects. The significance
of these differences is unclear and may be due at least in part to
variations in the amount and activity of the FGF protein applied.
Importantly, all of the FGFs act by initiating the normal cascade of
gene expression that leads to the formation of a limb. Interestingly,
an ectopic limb forms only when FGF is applied focally, as, for
example, by insertion of an FGF-soaked bead or a pellet of
Fgf-expressing cells. The outcome differs when FGF genes are
expressed more widely in the interlimb region of the chick (Mima et al.
1995; Vogel et al. 1996) or mouse (Abud et al. 1996) embryo. Although
such nonfocal ectopic FGF gene expression results in outgrowth of the
interlimb region mesoderm, sometimes in the form of multiple
protrusions along the flank of the embryo, the outgrowths eventually
regress and an ectopic limb is not formed. It should also be noted that
ectopic expression of FGF genes can adversely affect the development of
the normal limb, resulting in deletions or reductions and fusions,
mainly of the proximal bones, as well as the appearance of extra digits
(Vogel et al. 1996).
The ectopic limbs that form in response to focal FGF signaling are
often remarkably complete. The type of limb that develops is generally
dependent on the location of the FGF source. Sources placed close to
the prospective forelimb territory usually induce wings, whereas those
placed close to the prospective hindlimb usually induce legs (Cohn et
al. 1997). However, careful examination of these limbs has shown that
they are mosaics. Thus, the ectopic wings have some features of the leg
and vice versa (Ohuchi et al. 1998).
One of the most striking features of the FGF-induced ectopic limbs is
that their A-P polarity is opposite that of the normal limb. Thus, at
early stages of development, Shh RNA is detected on the
anterior rather than the posterior side of the ectopic limb bud, and
the pattern of digits that subsequently develops is likewise reversed.
One possible explanation for this observation is that within the
interlimb region there is an A to P gradient of decreasing competence
to express Shh (i.e., polarizing activity) (Hornbruch and
Wolpert 1991); therefore, Shh is more likely to be induced on
the anterior than on the posterior side of the nascent ectopic limb
bud. It is also noteworthy that the AER of the ectopic limbs always
forms at the D-V border of the interlimb region (Crossley et al. 1996;
Altabef et al. 1997), presumably because the same mechanism that is
used to form the AER of the normal limb is employed in the formation of
the ectopic limb bud AER.
Source and identity of the endogenous limb inducer
The experiments showing that FGF beads can induce limb formation in the interlimb region have provided insights into both the nature of the inducer and the availability of competence to respond to it. The finding that FGF protein is sufficient to induce the formation of a complete limb suggests that the endogenous inducer of limb development may likewise be a member of the FGF family. The observation that limbs can be induced to form along much of the length of the embryo suggests that competence to respond to the limb-inducing signal is widespread along the A-P axis and that the induction of normal limb formation at specific A-P levels may involve some mechanism for restricting the availability of the inducer.
Information about the stage at which limb induction occurs and the
tissue source of the inducer has been obtained using an assay that
involves grafting the prospective limb-forming region to an ectopic
site in a chick host embryo and then determining whether the graft can
develop autonomously into a limb (Kieny 1969; Kieny 1970; Pinot 1970;
Stephens et al. 1993). Although there are some discrepancies among the
results, they indicate that limb induction has occurred by about stages
13-15 (~20-25 somites), whereas the first morphological
manifestations of limb bud formation do not become apparent until
several hours later, at stage 17 (~30 somites). Prior to stages
13-14, a limb will form only when tissue medial to the prospective
limb territory is included in the graft. At stages 8-9, limbs will
form only when Hensen's node, somites, and intermediate mesoderm (IM)
are included in the graft. At slightly later stages (10-11), only the
somites and IM are required, and later still (stages 12-14) inclusion
of the IM is sufficient for limb formation by the graft. Thus, the
signal for limb induction appears to be transmitted in a medial to
lateral sequence.
Data from foil barrier (Stephens and McNulty 1981; Strecker and
Stephens 1983) and tissue ablation (Geduspan and Solursh 1992) experiments are consistent with the hypothesis that at stages 13-15,
the source of the limb inducer is the IM, which lies between the somite
and the lateral plate mesoderm and is comprised of nephrogenic mesoderm
(NM) and the Wolffian duct (WD) (Fig. 4C). Like all other tissues that
lie along the A-P body axis, the IM undergoes a series of
developmental changes in a rostral to caudal sequence. Initially it
consists of a small group of undifferentiated NM cells. As development
proceeds, the Wolffian duct forms and extends caudally, immediately
dorsal to the NM. The duct apparently produces signals that induce the
differentiation of the NM (Bishop-Calame 1965; Wolff 1970), which
condenses and forms the mesonephric tubules that subsequently fuse with
the duct to form the mesonephros, the principal organ of excretion
during early fetal life in many species. Eventually the mesonephros
degenerates and is functionally replaced by the metanephros (kidney).
Together, the data indicate that limb induction in the chick occurs
between stages 13 and 15, that the source of the inducer is most likely
the IM, and that a member of the FGF family is a good candidate for the
inducer. Moreover, the application of FGF protein to the interlimb
region induces limb development, which suggests that the endogenous
inducer is not normally present in the interlimb region. On the basis
of its expression pattern in the NM and its ability to induce ectopic
limb development, it was suggested that FGF8 is an endogenous inducer
of limb development, although the possibility that other FGF family
members may have a similar function was not excluded (Crossley et al.
1996). Furthermore, on the basis of expression patterns at earlier
stages, it was suggested that FGF8 may also be a component of the
limb-inducing signals that emanate from more medial tissues at earlier
stages (Vogel et al. 1996).
The expression data also raised the possibility that Fgf8
might be regulated by signals from the WD, suggesting a potential link
between the presumed limb inducer and development of the mesonephros
(Crossley et al. 1996). However, although experimental manipulations
that prevent the duct from extending caudally inhibit mesonephric
tubulogenesis, they do not prevent limb development (Le Douarin and
Fontaine 1970; Fernandez-Teran et al. 1997). Thus, if Fgf8
expression in the NM is induced by signals from the duct, it cannot be
involved in limb development. This issue was recently explored by
Fernandez-Teran et al. (1997), who studied Fgf8 expression in
embryos in which caudal extension of the WD was blocked at stages
9-11. They concluded that Fgf8 expression is induced and maintained by signals from the duct. However, results from a similar study by other investigators indicate that Fgf8 expression is detected in the absence of the duct in undifferentiated NM at the level
of the prospective forelimb between stages 13 and 15; at later stages
Fgf8 expression is not detected in the NM (J. Kulhman, G. Kardon, and L. Niswander, pers. comm.). Such data indicate that
Fgf8 is initially expressed in the NM irrespective of whether
the WD is present but that maintenance of this expression may depend on
signals from the duct. Although these results are consistent with the
hypothesis that FGF8 functions as the inducer of limb formation, this
idea must remain speculative until it can be rigorously tested, for
example, by inactivating the Fgf8 gene specifically in the NM
at a stage prior to limb induction.
Induction of Fgf8 expression in the prospective limb bud ectoderm
An early molecular manifestation of limb induction is the
expression of Fgf8 in the prospective limb ectoderm. If an FGF
produced in the NM is the limb inducer, then there is presumably some
mechanism for transmitting the inducing signal through the LPM to the
responding ectoderm (see Fig. 4C), as FGFs are thought to signal only
locally. It has been proposed that FGF10 may be the mesenchymal
mediator of the limb induction signal and thus function as the
endogenous inducer of Fgf8 expression in the overlying
ectoderm (Ohuchi et al. 1997). Several observations support this idea:
(1) In normal limb development, Fgf10 expression in the
mesoderm precedes that of Fgf8 in prospective limb ectoderm;
(2) when ectopic limbs are induced by FGF8, Fgf10 RNA is
detected in the mesenchyme before Fgf8 RNA is detected in the
overlying ectoderm; and (3) implants of rat Fgf10-expressing
cells in the interlimb region induce Fgf8 expression in the
overlying ectoderm.
Genetic evidence consistent with the hypothesis that FGF10 produced in
the prospective limb mesoderm induces Fgf8 expression in the
overlying ectoderm comes from studies of embryos homozygous for
Fgfr2IgIII, a presumed hypomorphic
allele of Fgfr2, in which limb development fails at the stage
of initial outgrowth (Xu et al. 1998). In the mutant embryos,
Fgf10 is expressed in its normal domain in the prospective
limb mesoderm but at significantly lower levels than in wild-type
embryos, and Fgf8 is not expressed in the prospective limb
ectoderm. According to the model proposed by Xu et al. (1998), the lack
of Fgf8 expression may be due to a loss of Fgfr2
function in the surface ectoderm, which prevents the ectodermal cells
from receiving the FGF10 signal that normally induces them to express Fgf8. Alternatively, it is possible that the Fgfr2
mutation acts at an earlier stage, interfering with transduction of the
limb-inducing signal through the prospective limb mesoderm and
resulting in the production of FGF10 at levels insufficient to induce
Fgf8 expression in the ectoderm. In either case, the absence
of Fgf8 expression in the ectoderm can account for the
observation that Fgf10 expression gradually diminishes in the
underlying mesoderm, presumably because of an interruption in the
proposed Fgf8-Fgf10 positive feedback loop. Further genetic
analysis should help to clarify the relationship between these two genes.
| |
Concluding remarks |
|---|
Because the purpose of this article is to review what is known
about the role of FGFs in vertebrate limb development, it seems appropriate to conclude with a diagram that summarizes the presumed functions of different FGF family members at various stages of limb
formation (Fig. 5). It should be apparent from the
preceding discussion that in most cases conclusive proof is lacking
that the specific FGFs indicated in the diagram perform the functions attributed to them, and that other members of the FGF family, either
known or yet to be identified, may be responsible for the functions
ascribed to a particular FGF. However, most of the existing hypotheses
about the functions of specific FGFs are testable and will soon be
evaluated, and other questions about FGF signaling, such as the
specificity of ligand-receptor interactions in vivo, the extent to
which FGF signals can act at a distance from the cells that produce
them, and the role of endogenous inhibitors of FGF signaling (Hacohen
et al. 1998) will presumably also be resolved. With the combination of
classic experimental embryological approaches and contemporary
techniques for analyzing gene function in the chick that is currently
available, as well as standard and sophisticated new methods of genetic
analysis in the mouse, knowledge about the molecular mechanism of
vertebrate limb development should continue to accumulate rapidly.
|
Finally, in view of the extraordinary extent to which developmental
mechanisms and the molecules that regulate them have been conserved
between invertebrates and vertebrates, it is somewhat surprising that
members of the FGF family have not yet been found to play a role in
Drosophila limb development. However, the recent identification of an FGF family member in Drosophila
(Sutherland et al. 1996), which plays a role in branching of the
tracheal system that may be analogous in some respects to the role
proposed for FGF10 in budding of vertebrate lung epithelium (Bellusci
et al. 1997), suggests that FGF function in the development of many different organs may have been conserved from invertebrates to vertebrates. As knowledge of the molecular mechanism of limb
development in Drosophila has informed and influenced many of
the recent studies of vertebrate limb development, it would be
gratifying if what is known about FGF function in vertebrate limb
development were to lead to new insights into the mechanism of
Drosophila limb development.
| |
Acknowledgments |
|---|
I am very grateful to the many people who provided unpublished information, preprints of manuscripts in press, and helpful discussions, and especially to P. Crossley, C. Logan, J. Michaud, L. Prentice, and G. Shoenwolf for providing photographs used in the figures. I would also like to thank M. Altabef, E. Coucouvanis, C. Loomis, L. Niswander, C. Tabin, C. Tickle, T. Vogt, and the members of my laboratory group for critically reading the manuscript and providing insightful suggestions for improving it.
| |
Footnotes |
|---|
1 E-MAIL gmartin{at}itsa.ucsf.edu; FAX (415) 476-3493.
| |
References |
|---|
|
Top
Introduction
References
|
|---|
