Neuron, Vol. 27, 251–263, August, 2000, Copyright 2000 by Cell Press
Midline Signals Regulate Retinal
Neurogenesis in Zebrafish
Ichiro Masai,*‡§ Derek L. Stemple,†
Hitoshi Okamoto,* and Stephen W. Wilson‡§
* Laboratory for Developmental Gene Regulation
Brain Science Institute
RIKEN (The Institute of Physical and
Chemical Research)
2-1 Hirosawa, Wako-shi
Saitama, 351-0198
Japan
†National Institute for Medical Research
The Ridgeway
Mill Hill
London NW7 1AA
‡ Department of Anatomy and Developmental
Biology
University College London
Gower Street
London WC1E 6BT
United Kingdom
Summary
In zebrafish, neuronal differentiation progresses across
the retina in a pattern that is reminiscent of the neurogenic wave that sweeps across the developing eye in
Drosophila. We show that expression of a zebrafish
homolog of Drosophila atonal, ath5, sweeps across
the eye predicting the wave of neuronal differentiation.
By analyzing the regulation of ath5 expression, we
have elucidated the mechanisms that regulate initiation and spread of neurogenesis in the retina. ath5
expression is lost in Nodal pathway mutant embryos
lacking axial tissues that include the prechordal plate.
A likely role for axial tissue is to induce optic stalk
cells that subsequently regulate ath5 expression. Our
results suggest that a series of inductive events, initiated from the prechordal plate and progressing from
the optic stalks, regulates the spread of neuronal differentiation across the zebrafish retina.
Introduction
During early forebrain development, the eye field initially
occupies a single domain in the anterior neural plate
(Woo and Fraser, 1995). Subsequent to this, cell movements within the midline of the neural plate, coupled
with signals derived from axial midline tissue lead to the
formation of two bilateral retinal primordia (Rubenstein
and Beachy, 1998; Varga et al., 1999). These primordia
evaginate from the brain, form optic cups, and differentiate as neural retina and pigment epithelium.
The Hedgehog (Hh) and Nodal families of secreted
signaling molecules are involved in splitting of the retinal
field. shh mutations cause severe cyclopia in mouse
and human (Chiang et al., 1996; Roessler et al., 1996)
but not in zebrafish (Schauerte et al., 1998; Karlstrom
§ To whom correspondence should be addressed (e-mail: imasai@
brain.riken.go.jp [I. M.], s.wilson@ucl.ac.uk [S. W. W.]).
et al., 1999). This suggests either that there is redundancy in this pathway in fish or that other factors contribute to the splitting of the retinal primordia. In support
of a role for Hh signaling in patterning the optic vesicles
in fish, there is the observation that overexpression of
Hh proteins suppresses retina and expands optic stalk,
a phenotype in some ways opposite to cyclopia (Ekker
et al., 1995; Macdonald et al., 1995).
The importance of Nodal signaling in early eye patterning is evident from mutations in this pathway in zebrafish. cyclops (cyc), squint (sqt), and one-eyed pinhead (oep) mutant embryos all show varying degrees of
cyclopia (Hatta et al., 1991; Heisenberg and NüssleinVolhard, 1997; Schier et al., 1997). cyc and sqt encode
Nodal ligands that have both discrete and partially redundant functions (Feldman et al., 1998; Rebagliati et
al., 1998; Sampath et al., 1998). Cyc is required for specification of ventral midline CNS tissue, and a failure in
migration of this tissue contributes to the failure of bilateral separation of the eyes in cyc⫺/⫺ embryos (Varga et
al., 1999). Sqt appears to act earlier in development than
Cyc and is required for prechordal plate development
(Feldman et al., 1998). Indeed, defects in signaling from
the prechordal plate may contribute to the eye defects
of sqt⫺/⫺ embryos (Heisenberg and Nüsslein-Volhard,
1997). Embryos lacking both Sqt and Cyc lack mesendoderm, revealing a partially redundant function for each
gene in mesoderm induction (Feldman et al., 1998,
2000). The oep gene encodes a EGF-CFC protein (Zhang
et al., 1998) required for Nodal signaling activity and
embryos missing zygotic Oep function lack prechordal
plate and exhibit a cyclopic phenotype similar to cyc
(Schier et al., 1997). Removal of both maternal and zygotic Oep function generates the same phenotype as
cyc;sqt double mutants, confirming that Nodal signaling
requires Oep function (Gritsman et al., 1999).
After formation of the optic cup, multipotent retinal
precursors give rise to all major cell types in the retina
(Harris, 1997). Neuronal differentiation is usually initiated
in the central retina and subsequently spreads peripherally (McCabe et al., 1999, and references within). In zebrafish, differentiation is initiated in ventronasal rather
than central retina, but, as in other vertebrates, neuronal
differentiation spreads as a wave from this position
(Raymond et al., 1995; Laessing and Stuermer, 1996;
Schmitt and Dowling, 1996; Hu and Easter, 1999). The
mechanisms underlying the progression of neurogenesis across the vertebrate retina are unknown but are
reminiscent of the neurogenic wave that sweeps across
the Drosophila eye imaginal disc generating ommatidia
behind the advancing morphogenetic furrow (Heberlein
and Moses, 1995). In this process, Hh secreted from
mature ommatidial neurons initiates neurogenesis in adjacent undifferentiated cells through induction of the
proneural bHLH transcription factor encoding gene
atonal. Newly differentiated neurons in turn secrete Hh
that initiates further neuronal differentiation and progression of the morphogenetic furrow. In this way, the
cycle of Hh secretion and Atonal activation underlies
the progression of neurogenesis across the eye disc.
In this study, we describe the expression and regulation of ath5, a zebrafish homolog of Drosophila atonal
that is exclusively transcribed in the developing retina
Neuron
252
Figure 1. ath5 Expression Predicts the Pattern of Neuronal Differentiation in the Embryonic Zebrafish Retina
(A–E) ath5. Lateral views of eyes at 25 hr (A), 27 hr (B), 30 hr (C), and 36 hr (D) and ventral view at 3 days (E). ath5 expression is initiated in
the ventronasal retina (closed arrowhead) adjacent to the choroid fissure (open arrowhead) and spreads throughout the retina in a wave-like
fashion. By 3 days, expression is localized to cells near the ciliary margin.
(F–J) islet-1. Expression (arrowhead) is initiated at 27 hr (expression at 25 hr is in the brain) and subsequently spreads through the retina
similar to ath5.
(K–O) lim3. Expression is initiated in ventronasal retina at 27 hr (arrowhead) and spreads through the retina similar to ath5.
(P–Q) neuroD. Expression is initiated in nasal retina (arrowhead) and spreads dorsally in the outer retina.
(R) Anti-tubulin antibody–labeled eye. Arrowhead shows early differentiating retinal ganglion cells in the ventronasal retina.
(S) FRet43-labeled eye. The earliest differentiating double cones (arrowhead) are located in ventronasal retina adjacent to the choroid fissure.
(T–V) Eyes labeled with ath5 RNA probe (blue) and anti-phosphorylated histone H3 antibody (brown). Arrowheads indicate dividing cells
throughout the retina at 25 hr and localized to the ciliary marginal zone at 33 hr and 48 hr. The arrow in (V) indicates one dividing cell within
the domain of ath5 expression. (T) shows a whole eye, and (U) and (V) show sections through the retina.
(W and X) Dorsal views of control and ath5-injected 2 somite stage and 70% epiboly stage embryos labeled to show neuroD (nrd) and huC
expression, respectively. There is robust induction of both genes in the ath5-injected embryos.
Abbreviations: cmz, ciliary marginal zone; dr, dorsal retina; inl, inner nuclear layer; nr, nasal retina; pl, photoreceptor cell layer; rgl, retinal
ganglion cell layer; tr, temporal retina.
in a pattern that predicts neuronal differentiation. ath5
expression is absent in oep mutants, suggesting that
axial tissues, possibly the prechordal plate, are required
to initiate ath5 expression and subsequent neuronal differentiation. However, transplant and explant experiments suggest that ath5 expression is not initiated
through direct interaction between prechordal plate and
prospective retina. We favor the possibility that prechordal plate influences the formation of optic stalk tissue that subsequently regulates ath5 expression in the
retina. In support of this, ath5 is always induced directly
adjacent to optic stalk tissue wherever this may be
within the optic vesicle; ablation of the interface between
retina and optic stalk inhibits ath5 expression; and transplanted optic stalk cells can induce ath5. We also show
that in the absence of induction of ath5 in ventronasal
retina, expression fails to progress to dorsal and temporal retina, suggesting that a retinal neurogenic wave may
be regulated by a series of short-range signaling events
that proceed across the retina. These results reveal unexpected similarities between the mechanisms underlying neurogenesis in the eyes of flies and vertebrates.
Retinal Neurogenesis in Zebrafish
253
Results
Zebrafish ath5 Shows Retina-Specific Expression
Closely Correlated with Neuronal Differentiation
In the developing fly eye, Atonal is essential for progression of the neurogenic wave and for neuronal differentiation (Jarman et al., 1994). To elucidate how neurogenesis
is regulated in the developing vertebrate retina, we isolated a zebrafish homolog of Drosophila atonal by cDNA
library screening. The bHLH motif of the zebrafish clone
has 87% amino acid identity to Xath5a (Kanekar et al.,
1997) and 68% to Drosophila Atonal (data not shown).
Phylogenetic tree analysis suggests that the zebrafish
gene is an ortholog of Xath5 (data not shown), and so we
designated this gene ath5. However, unlike the Xenopus
gene, in situ hybridization analysis revealed that ath5 is
expressed exclusively in the developing retina.
ath5 expression is first detected in the ventronasal
retina adjacent to the choroid fissure at 25hpf (Figure
1A), spreads from nasal to more dorsal and temporal
retina over the next few hours (Figures 1B and 1C), and
is present throughout most of the neural retina at 36 hr
(Figure 1D). By 72 hr, when retinal ganglion cell, inner
nuclear, and photoreceptor cell layers are distinct, ath5
expression is downregulated in differentiated neurons
but remains expressed in retinoblast cells near the ciliary
margin (Figure 1E).
The initiation and fan-shaped spread of ath5 expression are reminiscent of the pattern of appearance of
differentiated neurons at slightly later stages (Raymond
et al., 1995; Schmitt and Dowling, 1996; Hu and Easter,
1999). To examine the relationship between ath5 and
neuronal differentiation, we compared ath5 expression
to that of early neuronal markers. islet-1 (Korzh et al.,
1993; Inoue et al., 1994) and lim3 (Glasgow et al., 1997)
encode LIM homeodomain proteins that are expressed
in retinal ganglion and inner nuclear layer cells (Figures
1J and 1O). Expression of both genes is initiated in
ventronasal retina soon after ath5 and subsequently
spreads to dorsal and temporal retina, following the
progression of ath5 (Figures 1F–1O).
We confirmed that neuronal differentiation is indeed
initiated in ventronasal retina by examining other neuronal markers. neuroD encodes a bHLH protein expressed during late phases of neurogenesis (Korzh et al.,
1998). Within the retina, neuroD expression is initiated in
the ventronasal region and subsequently spreads dorsally and temporally within the outer retina (Figures 1P
and 1Q). Likewise, the first retinal ganglion cells to be
labeled with anti-acetylated ␣-tubulin antibody are present in the ventronasal retina (Figure 1R) as are the earliest FRet43-labeled double cone photoreceptors (Figure
1S; Larison and Bremiller, 1990).
We examined the relationship between ath5 expression and proliferation by using an antibody to phosphorylated histone H3 that labels cells in late G2 and M phase
(Wei and Allis, 1998). At the stage when ath5 expression
is initiated, the nuclei of dividing cells are present
throughout the retina near the scleral surface (Figure
1T). As ath5 expression spreads, dividing cells become
more localized to the margins of the retina (Figure 1U),
and by 48 hr, most dividing cells are located within
the ciliary marginal zone peripheral to, or occasionally
within, the domain of ath5 expression (Figure 1V). These
results indicate that ath5 is neither expressed in the
most proliferative cells of the ciliary marginal zone nor
within differentiated neurons of the central retina. Instead, in agreement with data in Xenopus on Xath5 (Perron et al., 1998), we suggest that ath5 is expressed
Figure 2. ath5 Expression Is Absent in oep⫺/⫺ Embryos but Present
in cyc⫺/⫺ Embryos
Lateral views of whole embryos.
(A and B) ath5 expression in cyc⫺/⫺ embryos. ath5 expression is
initiated and spreads throughout the cyc⫺/⫺ retina. The open arrowhead indicates the choroid fissure.
(C and D) ath5 expression in oep⫺/⫺ embryos. ath5 expression is
absent at both stages of development.
(E) Pax6 expression in an oep⫺/⫺ embryo. Pax6 is present in the
retina, lens, and diencephalon.
(F) rx1 expression in an oep⫺/⫺ embryo. rx1 is expressed throughout
the oep⫺/⫺ retina.
Abbreviations: d, diencephalon; l, lens; nr, nasal retina; rpe, retinal
pigment epithelium; tr, temporal retina.
transiently in retinoblasts and perhaps postmitotic neurons prior to full differentiation.
To determine if Ath5 can promote neurogenesis, we
injected DNA encoding Ath5 under the control of the
CMV promoter. Injected embryos exhibited very strong
induction of neuroD, a potential target for Ath5 within
the retina (Figure 1W). Supporting the possibility that
Ath5 may directly regulate neuroD expression, ectopic
ath5 often cell autonomously induced neuroD (data not
shown). Confirming that Ath5 could promote formation
of neurons, injected embryos also showed strong and
precocious expression of huC (Kim et al., 1996), an early
marker of postmitotic neurons (Figure 1X).
ath5 Expression and Retinal Differentiation Are
Disrupted in oep⫺/⫺ Embryos but Are
Relatively Normal in cyc⫺/⫺ Embryos
Mutations that affect the specification or migration of
axial midline tissue lead to cyclopic phenotypes in which
optic stalk tissue is variably reduced and retinal tissue
is fused across the midline (Brand et al., 1996; Varga et
al., 1999). To determine if the wave of retinal differentiation is affected in cyclopic mutants, we examined ath5
expression in oep⫺/⫺ and cyc⫺/⫺ embryos.
In cyc⫺/⫺ embryos, ath5 expression is initiated in a
single spot in nasal retina at the midline of the fused
eye (Figure 2A) and subsequently spreads bilaterally
from this position throughout the retina (Figure 2B). In
contrast to cyc, ath5 expression fails to be induced at
around 25hpf in oep⫺/⫺ embryos (Figure 2C) and is still
Neuron
254
Figure 3. Retinal Differentiation Is Disrupted in oep⫺/⫺ Embryos but Not cyc⫺/⫺ Embryos
(A–C) Plastic sections of 3 day wild-type (A), cyc⫺/⫺ (B), and oep⫺/⫺ (C) retinas. All retinal layers are present in the cyc⫺/⫺ retina, and retinal
ganglion cells form a thick axonal layer (arrowheads). In the oep⫺/⫺ retina, layering is severely disrupted although a thin outer nuclear layer
(asterisk) and pigment epithelium (inset) do appear to be present.
(D–F) islet-1 expression in 2 day wild-type (D), cyc⫺/⫺ (E), and oep⫺/⫺ (F) retinas. islet-1 expression is drastically reduced in the oep⫺/⫺ retina
but normal in the cyc⫺/⫺ retina.
(G–I) lim3 expression in 2 day wild-type (G), cyc⫺/⫺ (H), and oep⫺/⫺ (I) retinas. lim3 expression is retained in the cyc⫺/⫺ retina but absent in
the oep⫺/⫺ retina.
(J–L) rx1 expression in 2 day wild-type (J), cyc⫺/⫺ (K), and oep⫺/⫺ (L) retinas. rx1 expression is localized to ciliary marginal zone (arrowheads)
and outer layer in wild-type and cyc⫺/⫺ retina but is restricted to the marginal region of the oep⫺/⫺ retina.
(M–O) red opsin expression in 3 day wild-type (M), cyc⫺/⫺ (N), and oep⫺/⫺ (O) retinas. red opsin expression is observed in the outer layer in
the wild-type and cyc⫺/⫺ retinas but absent in the oep⫺/⫺ retina.
(P–R) GABA immunolabelling of cryosections of 3 day wild-type (P), cyc⫺/⫺ (Q), and oep⫺/⫺ (R) retinas. GABA expression is localized to the
retinal ganglion cell layer and amacrine cells in wild-type retinas. Weak GABA expression is observed in the cyc ⫺/⫺ retina (arrowheads) but
is absent in the oep ⫺/⫺ retina.
Except for the inset in (C), all embryos were treated with phenylthiourea (PTU) to reduce pigmentation.
Abbreviations: a, amacrine cells; g, retinal ganglion cell layer; ip, inner plexiform layer; in, inner nuclear layer; n, neural retina; o, outer nuclear
layer; pe, pigment epithelium.
absent at stages when ath5 is normally expressed
throughout the retina (Figure 2D). Therefore, although
cyc and oep exhibit superficially similar cyclopic defects, ath5 expression is only lost in oep mutants.
To determine if retinal identity was disturbed in oep
mutants, we examined expression of pax6 and rx1,
genes that are initially expressed throughout the retina
and later in proliferative cells of the ciliary marginal zone
(Macdonald and Wilson, 1997; Mathers et al., 1997;
Chuang et al., 1999). Expression of Pax6 and rx1 was
present throughout the retinas of oep⫺/⫺ eyes (Figures
2E and 2F), indicating that retinal identity is maintained
and suggesting that oep⫺/⫺ retinal cells remain undifferentiated and do not undergo normal neuronal differentiation.
To determine if the alterations in early retinal gene
expression correlate with later defects in retinal differentiation, we examined lamination and neuronal differentiation in oep⫺/⫺ and cyc⫺/⫺ eyes. In 3 day wild-type and
cyc⫺/⫺ retinas, lamination is advanced and nuclear and
plexiform layers have formed (Figures 3A and 3B; Fulwiler et al., 1997). In contrast, the laminar structure of
the oep⫺/⫺ retina is disrupted (Figure 3C). Although an
outer nuclear layer was variably present, cells throughout the rest of the retina were disorganized and very
little plexiform tissue was present.
Examination of markers of neuronal differentiation
demonstrated that retinal neurons are indeed reduced
in oep⫺/⫺ retinas. islet-1 and lim3 are expressed in retinal
ganglion and inner nuclear cells in both wild-type and
cyc⫺/⫺ retinas (Figures 3D, 3E, 3G, and 3H), whereas
expression of both genes is severely reduced or absent
in oep⫺/⫺ retinas (Figures 3F and 3I). The neurotransmitter GABA is expressed in the same cell layers by 3 days
in wild-type embryos (Sandell et al., 1994; Figure 3P).
GABA expression is weak but could be detected in
cyc⫺/⫺ retinas (Figure 3Q) but not in oep⫺/⫺ eyes (Figure
3R). rx1 is expressed in the proliferating margins and in
the outer nuclear layer at 2 days in wild-type (Chuang
et al., 1999) and cyc⫺/⫺ retinas (Figures 3J and 3K);
whereas, although weak expression is detected in the
ciliary margins of oep⫺/⫺ retinas, it is not observed in
the outer nuclear layer (Figure 3L). Red opsin is the first
opsin detected in differentiating retinal photoreceptors
(Raymond et al., 1995), and expression is present in
both wild-type and cyc⫺/⫺ eyes (Figures 3M and 3N)
Retinal Neurogenesis in Zebrafish
255
Table 1. Wild-Type Cells Can Rescue ath5 Expression in oep⫺/⫺ Embryos
Location of Transplanted Cells
Embryo
ath5
Telencephalon
Eye
Prechordal Plate Mesenchyme
Hatching Gland
Heart
1–3
1–6
3–1
3–5
4–1
4–4
1–1
1–2
1–4
1–5
3–2
3–3
3–4
4–2
4–3
4–5
⫹
⫹
⫹
⫹
⫹
⫹
⫺
⫺
⫺
⫺
⫺
⫺
⫺
⫺
⫺
⫺
⫹
⫹
⫺
⫹
⫹
⫹
⫹
⫹
⫹
⫺
⫹
⫹
⫹
⫺
⫹
⫹
⫹
⫺
⫺
⫺
⫺
⫺
⫹
⫺
⫺
⫺
⫹
⫺
⫹
⫺
⫺
⫺
⫹
⫹
⫹
⫹
⫹
⫹
⫺
⫺
⫺
⫹
⫺
⫺
⫺
⫹
⫺
⫺
⫹
⫹
⫹
⫺
⫹
⫹
⫺
⫺
⫺
⫺
⫺
⫺
⫺
⫹
⫺
⫺
⫺
⫺
⫺
⫺
⫹
⫺
⫺
⫺
⫺
⫺
⫹
⫺
⫺
⫺
⫺
⫺
Presence (⫹) or absence (⫺) of ath5 expression in the eyes of oep⫺/⫺ embryos in which wild-type cells were transplanted at late blastula
stage. The presence (⫹) or absence (⫺) of transplanted cells in the telencephalon, eye, prechordal plate mesenchyme, hatching gland, and
heart is indicated. The cyclopic phenotype was partially rescued in all embryos in which ath5 expression was rescued (1–3, 1–6, 3–1, 3–5,
4–1, and 4–4). Two embryos (1–5 and 4–2) in which a small number of wild-type cells were incorporated into the prechordal plate or hatching
gland showed no rescue of cyclopia or ath5 expression.
but is absent from oep⫺/⫺ eyes (Figure 3O). These data
indicate that, while a ciliary marginal zone is present
in the oep⫺/⫺ retina, most retinal cells fail to properly
differentiate.
To determine if the loss of expression of neuronal
markers in oep⫺/⫺ retinas could be due to extensive
early cell death, we monitored apoptosis from 24hpf to
2 days. Cell death was observed in the telencephalon
and lens of oep⫺/⫺ mutants, but relatively few apoptotic
profiles were detected at these stages in the retina of
wild-type, cyc⫺/⫺, or oep⫺/⫺ embryos (data not shown).
Together, these data indicate that the early absence
of ath5 expression is correlated with a later failure of
neuronal differentiation and lamination in the oep⫺/⫺ retina. Therefore, although the initial induction of retinal
tissue occurs normally in both cyc⫺/⫺ and oep⫺/⫺, later
retinal differentiation fails in oep⫺/⫺ embryos.
Prechordal Plate Tissue Regulates ath5 Expression
Although both cyc and oep mutations disrupt the Nodal
signaling pathway (Schier and Shen, 2000), they have
strikingly different retinal phenotypes. One difference
between cyc⫺/⫺ and oep⫺/⫺ is that prechordal plate tissue is retained in cyc⫺/⫺ embryos (Thisse et al., 1994),
whereas it is usually absent by the end of gastrulation
in oep⫺/⫺ embryos (Schier et al., 1997). This suggests
that spatial or temporal variations in residual Nodal signaling in cyc⫺/⫺ and oep⫺/⫺ embryos may lead to variable
retention or loss of anterior axial mesendoderm including prechordal plate. If this tissue subsequently influences the induction of ath5, then its presence/absence
could explain the differences in retinal development between cyc⫺/⫺ and oep⫺/⫺ embryos.
In support of this possibility, ath5 expression was only
lost in Nodal pathway mutant embryos that exhibited
severe axial defects. Thus, analysis of another Nodal
pathway mutation, squint (Feldman et al., 1998), showed
that ath5 was sometimes expressed, sometimes absent,
correlating with variable severity of axial deficits (data
not shown). Furthermore, retinal differentiation can occur in oep mutants (Malicki et al., 1996), and we have
also generated lines of oep fish that exhibit less severe
phenotypes and that retain ath5 expression. This suggests that allelic and genetic background differences in
oep fish alter the severity of the oep⫺/⫺ phenotype with
ath5 expression and retinal patterning only disrupted in
embryos exhibiting severe axial midline and cyclopic
defects.
To test whether axial tissues including the prechordal
plate tissue can indeed regulate ath5 expression and
retinal differentiation, we performed three sets of experiments. First, wild-type cells were transplanted into
oep⫺/⫺ embryos, and ath5 expression was assessed
with respect to the location of the transplanted cells.
Rescue of ath5 expression occurred in most cases when
transplanted cells were found in the hatching gland or
mesenchyme around the eyes (likely derivatives of the
prechordal plate), but not when cells were incorporated
into other tissues (Table 1; Figures 4A and 4B).
For the second set of experiments, we used a constitutively activated form of a putative Nodal/Activin serine/
threonine kinase receptor, TARAM-AD, which can rescue
mesendodermal tissue in oep⫺/⫺ embryos (Renucci et
al., 1996; Peyriéras et al., 1998). oep mutants in which
TARAM-AD was expressed showed ath5 transcripts in
midline retinal tissue similar to cyc⫺/⫺ embryos (Figure
4C; n ⫽ 14 of 15), suggesting that rescue of mesendodermal tissues rescues ath5 expression.
Finally, to determine if removal of the axial mesendoderm can phenocopy the oep retinal defects, we ablated
the embryonic shield, the source of axial mesendoderm.
Shield-ablated embryos showed varying degrees of
axial defects with more severely affected embryos lacking prechordal plate and notochord, and exhibiting an
oep-like fused eye (Figure 4D). In the retina of these
embryos (n ⫽ 6), ath5 transcripts were absent (Figure
4E), indicating that ath5 expression is dependent upon
organizer-derived tissues even in genetically wild-type
Neuron
256
Figure 4. Mesendoderm Regulates ath5 Expression
(A and B) An oep⫺/⫺ embryo into which dextran-labeled wild-type
cells (red in [A]) were transplanted. Transplanted cells have contributed to mesenchyme around the eye (A), and ath5 expression is
rescued (B).
(C) Lateral view of 27 hr oep⫺/⫺ embryo injected with RNA encoding
an activated form of TARAM-AD. This receptor rescues mesendoderm in oep⫺/⫺ embryos, and ath5 is present.
(D) Lateral view of a 24 hr embryo in which the embryonic shield
was removed. Axial mesendodermal derivatives including notochord (closed arrow) and hatching gland are absent, resulting in an
oep-like fused eye (open arrow).
(E) Lateral view of 33 hr shield-extirpated embryo labeled with an
ath5 RNA probe. No expression is detected in the fused eye.
retina. Together, these data suggest that loss of prechordal axial tissues underlies the defects in neurogenesis in the oep⫺/⫺ retinas.
The Influence of the Prechordal Plate upon ath5
Expression Is Indirect
Signals from prechordal mesendoderm could either directly act upon retinal tissue to induce ath5 expression
or could act upon other tissues that subsequently influence neurogenesis. To distinguish between these possibilities, we performed explant experiments to address
when signals from mesendoderm might be required.
When eyes (n ⫽ 8), or eyes attached to the brain (n ⫽
20), were dissected free of surrounding tissues at the
18 somite stage (7 hr before initiation of ath5 expression)
and cultured in vitro until 33 hr, ath5 was relatively normally induced (Figures 5A and 5B). This suggests that
signals from the prechordal mesendoderm are acting
earlier than 18 somite stage and are therefore unlikely to
directly influence ath5 expression. However, we cannot
exclude the possibility that a few non-CNS cells that we
failed to remove before culturing might be sufficient to
induce ath5.
ath5 expression is initiated in nasal retina and only
later spreads to temporal retina, suggesting that signals
that induce ath5 expression might initially be localized
to the nasal optic cup. To test this possibility, dissected
optic cups were divided into nasal and temporal halves
and were cultured separately. ath5 expression was observed only in nasal (n ⫽ 5 of 5; Figure 5C), but not
temporal (n ⫽ 7 of 7; Figure 5D), retinal explants. These
data suggest that inductive activity is localized to the
nasal side of the optic cup at the 18 somite stage.
ath5 Expression Is Initiated Adjacent to pax2.1Expressing Optic Stalk Tissue
ath5 expression is initiated in cells adjacent to the choroid fissure—the position at which the optic stalk connects with the retina—raising the possibility that this
Figure 5. ath5 Expression Is Closely Associated with pax2.1Expressing Optic Stalk Tissue
(A and B) ath5 expression in eyes or CNS tissue explanted at 18 hr
and cultured until 33 hr.
(C and D) Nasal (C) and temporal (D) retinal tissue explanted at 18
hr and cultured until 33 hr. ath5 expression is only observed in the
nasal retinal explant. The cultured retinal tissue curls around the
lens making the half eyes look similar to whole eyes.
(E) Wild-type eye, showing ath5 expression (blue) directly adjacent
to pax2.1 expression (red).
(F) Eye of a shh-injected embryo. Although the boundary between
optic stalk and retina is positioned more distally in the eye (arrowheads), ath5 expression (black) is still initiated directly adjacent to
the altered boundary of pax2.1 expression (red).
(G) Brain of a shh-injected oep⫺/⫺ embryo. pax2.1 (red) and ath5
(blue) are rescued in adjacent domains.
(H) Expression of pax2.1 in an oep⫺/⫺ brain. Expression is detected
at the midbrain/hindbrain boundary (arrow) but not in the rostral
forebrain.
tissue may be a source of signals that induce ath5. Optic
stalk cells and retinal cells that line the choroid fissure
express the paired box transcription factor Pax2.1/Noi
and differentiate as reticular astrocytes (Macdonald et
al., 1997). Analysis of ath5 and pax2.1 expression revealed that ath5-expressing retinal cells were directly
adjacent to pax2.1-expressing optic stalk/choroid fissure cells (Figure 5E), suggesting that interactions at
the boundary between optic stalk and neural retina may
initiate neurogenesis. To test this possibility, we experimentally manipulated the position of the optic stalk/
retinal interface.
Increasing Shh activity expands the optic stalk into
Retinal Neurogenesis in Zebrafish
257
Figure 6. The Boundary between Optic Stalk
and Neural Retina Can Regulate ath5 Expression
(A and B) Morphology of a normal eye (B) and
an operated eye (A) in an embryo in which the
optic stalk was ablated at 18 somite stage. On
the ablated side, the ventral retina remains
open (dots), whereas the choroid fissure has
closed (arrowheads) on the control side.
(C–E) Dorsal view of the head and dissected
eyes of embryos in which the optic stalk was
ablated ([C], left side, and [D]). ath5 expression is absent in the stalk-ablated eyes.
(F–H) Lateral views of control eye (H) and eyes
in which optic stalk tissue was ablated at the
18 somite stage (F and G). In the control eye,
pax2.1 is expressed at the choroid fissure and
ath5 throughout the retina. ath5 expression
is absent from the stalk-ablated eyes. Faint
residual pax2.1 expression is detected in
cells of the ventrotemporal choroid fissure/
optic stalk (arrow, [F]) and some ventral optic
stalk tissue is retained adjacent to temporal
retina in (G) (arrowheads).
(I–K) Dorsal view of the brain and dissected
eyes of an embryo in which dextran-labeled
optic stalk cells (brown) were transplanted
into the temporal retina (left side in [I]). ath5
expression is restricted to the ventronasal
retina in the control eye (K) but is ectopically
expressed in central retina (arrowheads) in
addition to ventronasal retina (arrow) in the
transplant eye (J).
(B), (E), (H), and (K) are printed in reverse.
Abbreviations: n, nasal retina; t, temporal
retina.
distal regions of the optic cup and represses retinal
fates, thereby repositioning the boundary between neural retina and optic stalk (Ekker et al., 1995; Macdonald
et al., 1995). In shh-overexpressing embryos, ath5 expression is always initiated directly adjacent to the most
distal pax2.1-expressing cells (Figure 5F). We also observed that at low efficiency, shh overexpression in
oep⫺/⫺ embryos led to induction of pax2.1 in a small
cluster of midline cells. In such oep⫺/⫺ embryos, ath5
expression was initiated in cells adjacent to the domain
of pax2.1 expression (Figure 5G). These data suggest
that the optic stalk or the boundary between neural
retina and optic stalk produces signals that induce ath5
expression. This is consistent with the observation that
a few pax2.1-expressing cells are present at the rostral
tip of the fused eye of cyc mutants (Macdonald et al.,
1995), whereas no pax2.1 expression is detected in the
oep⫺/⫺ eye (Figure 5H; Hammerschmidt et al., 1996).
Tissue at the Boundary between Optic Stalk and
Neural Retina Can Regulate ath5 Expression
To test the possibility that the optic stalk/retinal interface
induces ath5 expression, we ablated tissue at this
boundary either surgically or using a focused laser beam
(Figures 6A and 6B). When the ablation was performed
at 18hpf, ath5 expression was often absent in the retina
on the ablated side at 33hpf (Figures 6C and 6D; 10 of
25 embryos), whereas normal expression was observed
in control eyes (Figures 6C and 6E). Eyes in which ath5
expression was absent lacked pax2.1-expressing cells
directly adjacent to the nasal retina (Figures 6F–6H),
confirming that the boundary between optic stalk and
nasal retina is disrupted by the operation. These data
suggest that interaction between the optic stalk and
nasal retina is necessary for induction of ath5 expression. Furthermore, the absence of ath5 expression in
dorsal and temporal regions of eyes with optic stalk
ablations suggests that the spread of ath5 expression
to temporal retina is dependent upon initiation of expression in ventronasal retina.
To determine if optic stalk tissue can promote ath5
expression, optic stalk cells were transplanted into temporal retina. This was a difficult operation as cells often
failed to integrate into the retina. However, in some
cases, ectopic ath5 expression was observed adjacent
to the transplanted tissue (in three of nine embryos) at
a stage when ath5 expression is normally restricted to
ventronasal retina (Figures 6I–6K), suggesting that optic
stalk can cell nonautonomously promote ath5 expression in adjacent neural retina.
Nodal Signaling Inhibits pax2.1 and ath5 Expression
Our observations suggest that Nodal signaling mediated
by zygotic Oep induces prechordal mesendoderm that
promotes the formation of optic stalk tissue, which in
turn regulates retinal neurogenesis. However, further
analysis revealed that, conversely, Nodal signaling mediated by maternal and zygotic Oep may induce inhibitors of optic stalk fate/retinal neurogenesis, probably
through specification of nonaxial mesoderm.
Mesendodermal derivatives are reduced but still present to various degrees in cyc, oep, sqt, and shield-
Neuron
258
ablated embryos, and we have demonstrated that it is
loss of anterior axial mesendoderm that is correlated
with a failure to induce ath5. Nodal signaling is blocked
in maternal/zygotic oep mutants (MZoep), and in such
embryos there is a complete failure of induction and
involution of head and trunk mesendodermal derivatives
(Gritsman et al., 1999). Embryos that lack the activity of
both Cyc and Sqt similarly lack mesendoderm (Feldman
et al., 1998, 2000). To address the consequences of
complete blocking of Nodal signaling upon retinal neurogenesis, we examined pax2.1 and ath5 expression in
both MZoep embryos and in cyc;sqt double mutants.
Surprisingly, we found that pax2.1 expression recovers in MZoep mutants compared to zygotic oep mutants
(Figure 7A, compare to Figure 5H). Furthermore, ath5
expression is also present adjacent to the pax2.1 expression domain in MZoep mutants (Figure 7A). Similarly, ath5 is expressed in the anterior retina of sqt;cyc
double mutants (Figure 7B) despite the increased overall
phenotypic severity of double mutants compared to either single mutant alone. These experiments imply that
in zygotic oep⫺/⫺ embryos a residual degree of Nodal
signaling, dependent upon maternal Oep protein, inhibits the induction of pax2.1 and ath5 in eye tissue.
To confirm that blocking Nodal activity does lead to
reinitiation of pax2.1 and ath5 expression, we manipulated levels of Nodal signaling through overexpression
of a Nodal antagonist. Antivin is member of the Lefty
family of TGF␤ family proteins that can antagonize the
activity both of Nodals (Meno et al., 1999) and Activins
(Thisse and Thisse, 1999). Injection of 1 pg of antivin
RNA caused phenotypes that ranged in severity between zygotic oep and MZoep mutants (Figure 7C). In
embryos that exhibited phenotypes less severe than
MZoep or cyc;sqt double mutants, pax2.1 and ath5 expression was absent (Figures 7D and 7E). However,
higher doses of antivin led to reestablishment of both
pax2.1 and ath5 expression. Most remarkably, when
very high levels of antivin were expressed (Figure 7F),
pax2.1 was strongly induced in a narrow band of cells
at the boundary of the eye with the brain and ath5 was
expressed widely throughout the enlarged retina (Figures 7G and 7H).
These results show that a reduction in Nodal signaling
leads to loss of prechordal axial tissues and inhibition
of retinal neurogenesis, while complete loss of Nodal
signaling leads to the additional loss of other mesendodermal tissues and reinitiation of neurogenesis. One interpretation of these results is that mesendodermal tissues retained in zygotic Oep mutants inhibit retinal
neurogenesis and that prechordal axial tissue normally
relieves this inhibition (Figure 8C; see Discussion).
Discussion
In this study, we have shown that ath5, a zebrafish gene
homologous to Drosophila atonal, is expressed exclusively in the retina with a pattern closely linked to neuronal differentiation. By using ath5 as a marker for retinal
neurogenesis, we found that oep mutants lack ath5 expression and exhibit severely disturbed retinal differentiation. Cells in oep⫺/⫺ eyes do express early retinal markers, suggesting that the defect lies in differentiation of
retinal neurons rather than in the specification of retinal
identity. This suggests that specification of retinal fate
is not sufficient to initiate retinal neurogenesis and that
Figure 7. Depletion of Nodal/Activin Signaling Rescues pax2.1 and
ath5 Expression
(A) ath5 (blue) and pax2.1 (red) expression in an MZoep embryo.
ath5 expression (closed arrowheads) is present adjacent to pax2.1
expression (open arrowhead) at the midline of the eye.
(B) ath5 expression in a cyc;sqt double mutant embryo. ath5 is
expressed at the rostal tip of the eye (arrowheads).
(C–E) Whole embryo and eyes of embryos injected with 1 pg antivin
RNA. Embryos exhibit a phenotype of severity between zygotic and
maternal/zygotic oep mutants (C). Expression of both pax2.1 (D)
and ath5 (E) is absent from the cyclopic eyes.
(F–H) Whole embryo and eyes of embryos injected with more than
50 pg antivin RNA. Injected embryos have a large eye and brain
sitting on the yolk with only rudimentary trunk and tail (arrow) structures (F). pax2.1 expression is present as a narrow band of cells at
the interface between the eye and the brain (arrowheads, [G]). ath5
expression is widely expressed in the retina (H).
additional signals are required for cells to progress from
proliferation to differentiation.
Our studies suggest that extrinsic signals required for
initiation of retinal differentiation are indirectly dependent upon prechordal axial tissue. We observe a tight
correlation between induction of ath5 expression and
the presence of optic stalk tissue, suggesting that prechordal axial tissue may directly or indirectly induce
optic stalk tissue that is subsequently required for induction of retinal neurogenesis (Figure 8A).
Axial Tissues Regulate Retinal Neurogenesis
The prechordal plate has long been known to be an
important regulator of eye patterning (Adelmann, 1930).
Retinal Neurogenesis in Zebrafish
259
Figure 8. Interactions Potentially Regulating Retinal Neurogenesis
(A) Signals promoting optic stalk and retinal differentiation. Schematic of the brain, optic stalk (red), and retina (blue) with underlying
prechordal mesendoderm (dark green). Signals from the mesendoderm may directly (blue arrows) or indirectly (green arrow) promote
optic stalk fate. One indirect route may be through the induction
(green arrow) of hypothalamus (pale green), which in turn promotes
optic stalk fates (blue arrow). Signals (pale blue arrow) from the
telencephalon (yellow) may also promote optic stalk fates (Shanmugalingam et al., 2000). Signals (white arrow) from the optic stalk
may subsequently induce ath5 (dark blue) expression in ventronasal
retina. ath5 expression then spreads across the retina (black
arrows). The signaling events indicated are likely to be initiated at
stages earlier than illustrated.
(B) Sequential stages in the induction and spread of ath5. Illustrations of the neuroepithelium at the interface between the optic stalk
(red) and the neural retina (blue). Signals (arrows, top panel) from
the optic stalk may induce ath5 (dark blue, middle panel) in adjacent
retinal cells. These ath5-positive cells may subequently promote
(arrows, middle panel) ath5 expression in adjacent retinal cells,
thereby propagating ath5 expression across the retina (bottom
panel).
(C) Differences in optic stalk/eye development between wild-type
(WT), zygotic oep (oep), and MZoep and antivin-injected embryos.
Schematics of dorsal views of gastrula stage embryos. Prechordal
mesendoderm is dark green, posterior axial mesendoderm is maroon, and nonaxial mesendoderm is pink; prospective telencephalon
is yellow, prospective eye tissue is blue, prospective optic stalk is
red, neural and nonneural ectoderm is pale green, and the yolk cell
is black. The relative positions of neural derivatives are simplified
to reflect their approximate positions at the end of gastrulation (e.g.,
Varga et al., 1999). In the wild-type situation, prechordal mesendoderm may produce signals (⫹) that promote optic stalk fate by
antagonizing signals (⫺) from other mesendodermal derivatives.
Prospective telencephalon may also produce signals (⫹) that promote optic stalk fate. In the zygotic oep mutant, prechordal plate
is transformed to notochord (Gritsman et al., 2000), and other mesendodermally derived signals (⫺) that inhibit optic stalk fate predominate. However, in the MZoep and antivin-injected embryo, all head
and trunk mesendoderm is absent (Gritsman et al., 1999; Feldman
et al., 2000), and so inhibition of optic stalk fate is relieved.
It is required for the normal development of ventral CNS
and for splitting the initially single eye field into two
bilateral retinas (Li et al., 1997; Pera and Kessel, 1997).
Here we show that axial tissues, including prechordal
plate, appear to have at least two distinct and separable
roles in retinal development. The first of these is to mediate the separation of the two retinas, and the second is
to provide signals that initiate neurogenesis within each
retina. These two roles are distinguished by the cyc and
oep mutations, both of which affect the development of
axial midline tissues. In both oep⫺/⫺ and cyc⫺/⫺ embryos,
separation of the retinal fields is blocked; however, neurogenesis is only affected in oep mutants. Thus, the
signals that are required to induce ath5 and initiate retinal neurogenesis are present in cyc⫺/⫺ but are absent
in oep⫺/⫺ embryos.
The role of prechordal plate tissue in regulating eye
patterning is often interpreted as being a provider of
signals that divert midline cells away from a retinal fate,
thereby splitting the retinal field (Li et al., 1997). While
this may be in part true, analysis of zebrafish mutants
has shown that midline neurectodermal tissues also migrate anteriorly, thereby physically separating the retinal
fields, in addition to providing signals that alter medial
cell fates (Heisenberg and Nüsslein-Volhard, 1997; Marlow et al., 1998; Varga et al., 1999).
The anterior migration of midline neural plate cells
into the retinal field is blocked in cyc mutants (Varga et
al., 1999), and so differences in migration of midline
neural cells are unlikely to explain the retinal phenotypic
differences between cyc⫺/⫺ and oep⫺/⫺ embryos. Instead, prechordal plate tissue is retained in cyc⫺/⫺ embryos but lost from oep⫺/⫺ embryos, suggesting that
vertical signals from prechordal plate tissue may be
sufficient to maintain neurogenesis in cyc⫺/⫺ retina. In
oep⫺/⫺ embryos, the failure of axial cell movements coupled with the lack of signaling from prechordal plate
may in combination lead to formation of a cyclopic retina
within which neurogenesis fails to proceed.
oep⫺/⫺ embryos lack endoderm (Schier et al., 1997),
and the oep gene is broadly expressed in anterior ectoderm (Zhang et al., 1998), raising the possibility that
defects other than in prechordal plate signaling could
also contribute to the retinal neurogenesis phenotype.
However, several observations suggest that axial mesendodermal defects are the primary cause of the phenotype. First, physical removal of the source of axial mesendoderm can phenocopy the retinal oep defects,
indicating that the genetic status of the neural tissue is
not critical. Second, ath5 expression is rescued in
oep⫺/⫺ embryos when wild-type cells incorporate into
hatching gland and mesenchymal cells near the eye,
both likely to derive from mesendoderm at or close to
the midline.
The Role of Prechordal Plate in Retinal
Neurogenesis Is Likely to Be Indirect
Prechordal plate signals could influence ath5 expression either by directly acting upon retinal cells or by
indirectly affecting the development of other tissues that
subsequently affect retinal neurogenesis. The latter is
more likely as ath5 expression is retained after removal
of tissues around the brain 7 hr before the gene is first
expressed, suggesting that prechordal plate signals are
required early in development. Furthermore, ath5 expression is initiated in the complete absence of mesendoderm in MZoep, cyc;sqt, and antivin-injected embryos, implying that some other tissue must regulate
ath5 expression in these embryos.
Nasal optic stalk tissue may provide the signals that
regulate neurogenesis. First, in wild-type, shh-injected,
Neuron
260
cyc⫺/⫺, MZoep, and antivin-injected embryos, ath5 is
always expressed adjacent to pax2.1-expressing optic
stalk cells, wherever these may be located. Furthermore,
the loss of ath5 expression in oep⫺/⫺ embryos is correlated with an absence of optic stalk cells, physical ablation of optic stalk tissue can also result in a loss of ath5
expression, and, finally, transplanted optic stalk cells
can promote ath5 expression. Together, these data suggest that communication between optic stalk and neural
retina is required for induction of retinal neurogenesis
and that the role of the prechordal plate may be to
directly or indirectly promote optic stalk development
(Figure 8A).
Nodal Signaling May Induce Both Tissues/Signals
that Promote Retinal Neurogenesis
and Tissues/Signals that Inhibit
Retinal Neurogenesis
One striking observation is that although reduced Nodal
signaling in zygotic oep mutants is associated with a
loss of pax2.1 and ath5 expression, more complete loss
of Nodal signaling (MZoep, cyc;sqt, or antivin-injected
embryos) restores pax2.1/ath5. This indicates that in
zygotic oep⫺/⫺ embryos, residual Nodal activity inhibits
pax2.1/ath5 expression. As oep⫺/⫺ and MZoep embryos
both lack anterior axial mesendoderm including prechordal plate, then their contrasting eye phenotypes are
unlikely to be due to differences in this tissue. Instead,
the inhibition of ath5 and pax2.1 expression is likely to
be dependent upon other tissues that are present in
zygotic oep mutants but absent in MZoep mutants (Figure 8C).
Complete loss of Nodal signaling leads to a failure in
induction and involution of all head and trunk mesoderm
and endoderm (Schier and Shen, 2000). Inhibition of
ath5/pax2.1 expression could therefore be dependent
upon signals from posterior axial tissues or paraxial
mesoderm, both of which are present in zygotic oep
mutants but are absent in MZoep mutants. If the prechordal mesendoderm normally antagonizes the activity
of other mesendodermally derived signals, then its role
would become redundant and pax2.1/ath5 expression
would be restored in the absence of all mesendoderm
in Nodal-deficient embryos.
One consequence of the loss of all mesendoderm is
anteriorization of the ectoderm such that prospective
telencephalon (Gritsman et al., 1999) and eye (this study;
Thisse et al., 2000) are both enlarged. This may be due
to the loss of posteriorizing signals normally derived
from nonaxial mesendoderm. pax2.1 expression in the
optic stalk is in part dependent upon fgf8, most likely
derived from adjacent telencephalic cells (Shanmugalingam et al., 2000), suggesting that both axial midline
signals and anterior neural plate signals may act in concert to promote optic stalk fate. The enlargement of
anterior neural plate domains in embryos lacking Nodal
activity may thus provide an alternative source of signals
that rescue pax2.1 and ath5 expression (Figure 8C).
Optic Stalk Tissue May Act Nonautonomously
to Regulate Retinal Neurogenesis
ath5 expression is initiated adjacent to pax2.1-expressing cells that appear to be both required and sufficient to
promote ath5 expression. Although pax2.1 expression is
closely associated with ath5, Pax2.1 itself is not required
for the regulation of retinal neurogenesis. noi⫺/⫺ embryos carrying a null mutation in the pax2.1 gene exhibit
a failure in closure of the choroid fissure but exhibit
apparently normal ath5 expression and retinal differentiation (Macdonald et al., 1997; data not shown). Furthermore, ectopic expression of pax2.1 in the dorsal retina
does not lead to induction of ath5 expression (data not
shown). Mutations affecting Fgf8 signaling also affect
the development of the optic stalk, yet ath5 expression
and retinal differentiation are again not very severely
affected (Heisenberg et al., 1999; Picker et al., 1999;
Shanmugalingam et al., 2000; data not shown). Both
Pax2.1 and Fgf8 are members of large gene families,
and it remains a possibility that they act in concert with
other related proteins to regulate retinal neurogenesis.
Indeed, in vitro, Fgfs can promote retinal neurogenesis
(McFarlane et al., 1998), and further analyses will be
required to determine if they have comparable roles in
vivo.
Mechanisms Regulating Neurogenesis in the Eyes
of Vertebrates and Flies May Be Conserved
In zebrafish, ath5 expression and neuronal differentiation are initiated ventronasally and subsequently spread
to dorsal and temporal retina (this study; Raymond et
al., 1995; Laessing and Stuermer, 1996; Schmitt and
Dowling, 1996). Indeed, successive waves of neurogenesis spreading from nasal to temporal retina generate
retinal ganglion and amacrine cells, bipolar and horizontal cells, and finally photoreceptors (Hu and Easter,
1999). These waves of neurogenesis are reminiscent of
the neurogenic wave that occurs as the morphogenetic
furrow sweeps across the imaginal disc generating neurons in the developing Drosophila eye.
Regulation of the neurogenic wave in the eye of Drosophila is based upon cycles of non-cell-autonomous
activation of atonal by Hh and cell-autonomous activation of hh by Atonal (Heberlein and Moses, 1995). ath5
is the closest vertebrate homolog of atonal, and, as
ath5 predicts the wave of neuronal differentiation, it is
tempting to speculate that the zebrafish gene may play
a role comparable to its fly counterpart. Although
Hedgehog proteins may play a role in retinal development (Levine et al., 1997), as yet it is unclear if a wave
of Hh activity spreads across the vertebrate retina.
The neurogenic wave in the fly eye is regulated by a
successive relay of short-range signaling events, and
our studies suggest that ath5 expression in dorsal and
temporal retina is dependent upon initiation of ath5 expression in ventronasal retina. This raises the possibility
that a similar relay of short-range signals is involved
in the progression of retinal neurogenesis in zebrafish
(Figure 8B). However, at this stage, we cannot exclude
the possibility that signals from the optic stalk or ventronasal retina act over a long distance to influence neurogenesis.
Waves of retinal differentiation have been reported in
vertebrates other than fish, with neurogenesis generally
proceeding from central to peripheral retina (e.g., Prada
et al., 1991). However, it has recently been shown that
the progression of retinal ganglion cell differentiation
in the chick retina does not depend upon the presence
of earlier differentiated neurons, suggesting that the
spread of neurogenesis is regulated by a cell intrinsic
mechanism (McCabe et al., 1999). However, in fish, the
wave of ath5 expression occurs much earlier than any
Retinal Neurogenesis in Zebrafish
261
waves of neuronal differentiation, suggesting that inductive signals may already have spread throughout the
retina by the stage that the first neurons differentiate.
Further elucidation of the molecular pathways underlying the regulation of retinal neurogenesis will undoubtedly reveal the extent of conservation of these pathways
between species.
Experimental Procedures
Fish
Zebrafish (Danio rerio) were maintained at 28.5⬚C on a 14 hr light/
10 hr dark cycle, and embryos were collected by natural spawning.
oeptz57,oepz1, cyctf291, and cycb16 alleles were used (Hatta et al., 1991;
Brand et al., 1996; Hammerschmidt et al., 1996; Schier et al., 1997).
Isolation of ath5 and In Situ Hybridization
A zebrafish somite stage cDNA library (a gift of Dr. Grunwald) was
screened at low stringency with Xath5 using standard procedures.
cDNA clones were subcloned into pBluescript II SK⫹ (Stratagene)
and sequenced. In situ hybridizations with ath5 and other probes
were carried out using standard procedures (Macdonald et al.,
1995).
Antibody Labeling
Standard procedures were used (Wilson et al., 1990) with anti-Pax6
(Macdonald et al., 1995) at 1:400; FRet43 antibody (Larison and
Bremiller, 1990) at 1:10; anti-acetylated ␣ tubulin antibody (Sigma)
at 1:1000; anti-GABA (Sigma) at 1:500; and anti-phosphorylated histone antibody (Upstate biotechnology) at 1:500.
Sectioning
Embryos incubated in 0.003% phenylthiourea to inhibit pigmentation were fixed, dehydrated to 100% ethanol, embedded in JB4
resin (Agar Scientific Ltd), and sectioned (5 ␮m–10 ␮m) using a
tungsten knife on a Jung 2055 Autocut. For cryosectioning, fixed
embryos were transferred to 20% sucrose in phosphate buffer,
mounted in OCT compound, and sectioned at 8 ␮m on a Microm
cryotome.
Tissue Ablations
For shield ablations, embryos at 50% epiboly stage were dechorionated, washed with 1/3⫻ Ringer’s solution, and transferred to embryo-sized holes in agar. Glass capillaries pulled to fine tips were
broken with a razor blade such that the diameter of the tip corresponded to the size of the embryonic shield. The capillary needles
were filled with oil and placed on the shield, and slight suction
was applied until the shield region was removed. Shield-extirpated
embryos were incubated in 1⫻ Ringer’s solution for 30 min after
the operation, then transferred to 1/3⫻ Ringer’s solution at 28.5⬚C
until fixation.
For optic stalk ablations, 18 somite embryos were mounted in
3% methyl cellulose in 1⫻ Ringer’s solution under compound microscope optics. Cells at the boundary between optic stalk and neural
retina were ablated by laser (Micropoint Ablation Laser System, Dye:
Coumarin 440, Zeiss; N2 laser: YSL337ND, Laser Science, Inc.) or
with tungsten needles, after which embryos were left to develop in
1⫻ Ringer’s solution.
Cell Transplantation
Cells from wild-type embryos injected with rhodamine-dextran (5%
in 100 mM KCl) were transplanted into late blastula stage oep⫺/⫺
embryos. Locations of the transplanted cells were determined by
confocal laser scanning microscopy at 24 hr (LSM 510, Zeiss), and
embryos in which wild-type cells were incorporated into retina, telencephalon, hatching gland, prechordal plate, or heart were selected for further analysis. Embryos were fixed at 33 hr, and ath5
expression was analyzed.
For optic stalk transplants, donor embryos were labeled with a
mixture of rhodamine–dextran (5% in 100 mM KCl) and biotin-dextran (3%). Embryos at 20 hr of development were dissected, and
optic stalk tissue was isolated with tungsten needles. Dissected
optic stalk tissue was transplanted into the caudal retina of host
embryos using glass capillary needles. Transplanted embryos were
fixed at 25 hr and analyzed for ath5 expression and the location of
biotin (Vectastain detection kit).
In Vitro Culture of Dissected Eyes
Embryos at 18 somite stage were dissected in L15 medium with
tungsten needles. After separation of the head, the skin, prechordal
plate, and mesenchyme were removed from the brain and eyes. In
some cases, optic cups were isolated from the diencephalon by
cutting the optic stalk. Dissected eyes and eyes attached to forebrains were incubated at 28.5⬚C in L15 medium without serum, fixed
at 33 hr, and labeled with an RNA probe to ath5.
Acknowledgments
We thank Frédéric Rosa for providing TARAM-A-injected oep embryos, Monica Vetter for providing Xath5, Alex Schier and Will Talbot
for providing embryos, and other colleagues for providing reagents.
This study was supported by funds from the Ministry of Education,
Science, and Culture of Japan to I. M., a RIKEN BSI Grant (no. 116)
to H. O., and grants from the Wellcome Trust, EC, and BBSRC to
S. W. W., who is a Wellcome Trust Senior Research Fellow.
Received February 10, 2000; revised June 15, 2000.
References
Adelmann, H.B. (1930). Experimental studies on the development
of the eye. III. The effect of the substrate (“Unterlagerung”) on the
heterotopic development of median and lateral stripes of the anterior
end of the neural plate and amblystoma. J. Exp. Zool. 57, 223–281.
Brand, M., Heisenberg, C.-P., Warga, R.M., Pelegri, F., Karlstrom,
R.O., Beuchle, D., Picker, A., Jiang, Y.-J., Furutani-Seiki, M., van
Eeden, F.J.M., et al. (1996). Mutations affecting development of the
midline and general body shape during zebrafish embryogenesis.
Development 123, 129–142.
Chiang, C., Litingtung, Y., Lee, E., Young, K.E., Corden, J.L., Westphal, H., and Beachy, P.A. (1996). Cyclopia and defective axial patterning in mice lacking Sonic hedgehog gene function. Nature 383,
407–413.
Chuang, J.C., Mathers, P.H., and Raymond, P.A. (1999). Expression
of three Rx homeobox genes in embryonic and adult zebrafish.
Mech. Dev. 84, 195–198.
Ekker, S.C., Ungar, A.R., Greenstein, P., von Kessler, D.P., Porter,
J.A., Moon, R.T., and Beachy, P.A. (1995). Patterning activities of
vertebrate hedgehog proteins in the developing eye and brain. Curr.
Biol. 5, 944–955.
Feldman, B., Gates, M.A., Egan, E.S., Dougan, S.T., Rennebeck,
G., Sirotkin, H.I., Schier, A.F., and Talbot, W.S. (1998). Zebrafish
organizer development and germ-layer formation require nodalrelated signals. Nature 395, 181–185.
Feldman, B., Dougan, S.T., Schier, A.F., and Talbot, W.S. (2000).
Nodal-related signals establish mesendodermal fate and trunk neural identity in zebrafish. Curr. Biol. 10, 531–534.
Fulwiler, C., Schmitt, E.A., Kim, J.M.-J., and Dowling, J.E. (1997).
Retinal patterning in the zebrafish mutant cyclops. J. Comp. Neurol.
381, 449–460.
Glasgow, E., Karavanov, A.A., and Dawid, I.B. (1997). Neuronal and
neuroendocrine expression of lim3, a LIM class homeobox gene, is
altered in mutant zebrafish with axial signaling defects. Dev. Biol.
192, 405–419.
Gritsman, K., Zhang, J., Cheng, S., Heckscher, E., Talbot, W.S.,
and Schier, A.F. (1999). The EGF-CFC protein One-eyed pinhead is
essential for Nodal signaling. Cell 97, 121–132.
Gritsman, K., Talbot, W.S., and Schier, A.F. (2000). Nodal signalling
patterns the organiser. Development 127, 921–932.
Hammerschmidt, M., Pelegri, F., Mullins, M.C., Kane, D.A., Brand,
M., van Eeden, F.J.M., Furutani-Seiki, M., Granato, M., Haffter, P.,
Heisenberg, C.-P., et al. (1996). Mutations affecting morphogenesis
Neuron
262
during gastrulation and tail formation in the zebrafish, Danio rerio.
Development 123, 143–151.
Harris, W.A. (1997). Cellular diversification in the vertebrate retina.
Curr. Opin. Genet. Dev. 7, 651–658.
Hatta, K., Kimmel, C.B., Ho, R.K., and Walker, C. (1991). The cyclops
mutation blocks specification of the floor plate of the zebrafish
central nervous system. Nature 350, 339–341.
Heberlein, U., and Moses, K. (1995). Mechanisms of Drosophila retinal morphogenesis: the virtues of being progressive. Cell 81,
987–990.
Heisenberg, C.-P., and Nüsslein-Volhard, C. (1997). The function of
silberblick in the positioning of the eye anlage in the zebrafish embryo. Dev. Biol. 184, 85–94.
Heisenberg, C.-P., Brennan, C., and Wilson, S.W. (1999). Zebrafish
aussicht mutants exhibit widespread overexpression of ace (fgf8)
and coincident defects in CNS development. Development 126,
2129–2140.
Hu, M., and Easter, S.S., Jr. (1999). Retinal neurogenesis: the formation of the initial central patch of postmitotic cells. Dev. Biol. 207,
309–321.
Inoue, A., Takahashi, M., Hatta, K., Hotta, Y., and Okamoto, H. (1994).
Developmental regulation of Islet-1 mRNA expression during neuronal differentiation in embryonic zebrafish. Dev. Dyn. 199, 1–11.
Jarman, A.P., Grell, E.H., Ackerman, L., Jan, L.Y., and Jan, Y.N.
(1994). atonal is the proneural gene for Drosophila photoreceptors.
Nature 369, 398–400.
Kanekar, S., Perron, M., Dorsky, R., Harris, W.A., Jan, L.Y., Jan.,
Y.N., and Vetter, M.L. (1997). Xath5 participates in a network of
bHLH genes in the developing Xenopus retina. Neuron 19, 981–994.
Karlstrom, R.O., Talbot, W.S., and Schier, A.F. (1999). Comparative
synteny cloning of zebrafish you-too: mutations in the Hedgehog
target gli2 affect ventral forebrain patterning. Genes Dev. 15,
388–393.
Kim, C.-H., Ueshima, E., Muraoka,O., Tanaka, H., Yeo, S.-Y., Huh,
T.-L., and Miki, N. (1996) Zebrafish elav/HuC homologue as a very
early neuronal marker. Neurosci. Lett. 216, 109–112.
Korzh, V., Edlund, T., and Thor, S. (1993). Zebrafish primary neurons
initiate expression of the LIM homeodomain protein Isl-1 at the end
of gastrulation. Development 118, 417–425.
Korzh, V., Sleptsova, I., Liao, J., He, J., and Gong, Z. (1998). Expression of zebrafish bHLH genes ngn1 and nrd defines distinct stages
of neuronal differentiation. Dev. Dyn. 213, 92–104.
Laessing, U., and Stuermer, C.A. (1996). Spatiotemporal pattern of
retinal ganglion cell differentiation revealed by the expression of
neuolin in embryonic zebrafish. J. Neurobiol. 29, 65–74.
Larison, K.D., and Bremiller, R. (1990). Early onset of phenotype and
cell patterning in the embryonic zebrafish retina. Development 109,
567–576.
Levine, E.M., Roelink, H., Turner, J., and Reh, T.A. (1997). Sonic
hedgehog promotes rod photoreceptor differentiation in mammalian
retinal cells in vitro. J. Neurosci. 17, 6277–6288.
Marlow, F., Zwartkruis, F., Malicki, J., Neuhauss, S.C.F., Abbas, L.,
Weaver, M., Driever, W., and Solnica-Krezel, L. (1998). Functional
interactions of genes mediating convergent extension, knypek and
trilobite, during the partitioning of the eye primordium in zebrafish.
Dev. Biol. 203, 382–399.
Mathers, P.H., Grinberg, A., Mahon, K.A., and Jamrich, M. (1997).
The Rx homeobox gene is essential for vertebrate eye development.
Nature 387, 603–607.
McCabe, K.L., Gunther, E.C., and Reh, T.A. (1999). The development
of the pattern of retinal ganglion cells in the chick retina: mechanisms that control differentiation. Development 126, 5713–5724.
McFarlane, S., Zuber, M.E., and Holt, C.E. (1998). A role for the
fibroblast growth factor receptor in cell fate decisions in the developing vertebrate retina. Development 125, 3967–3975.
Meno, C., Gritsman, K., Ohishi, S., Ohfuji, Y., Heckscher, E., Mochida, K., Shimono, A., Kondoh, H., Talbot, W.S., Robertson, E.J., et
al. (1999). Mouse Lefty2 and zebrafish antivin are feedback inhibitors
of nodal signaling during vertebrate gastrulation. Mol. Cell 4,
287–298.
Pera, E.M., and Kessel, M. (1997). Patterning of the chick forebrain
anlage by the prechordal plate. Development 124, 4153–4162.
Perron, M., Kanekar, S., Vetter, M.L., and Harris, W.A. (1998). The
genetic sequence of retinal development in the ciliary margin of the
Xenopus eye. Dev. Biol. 199, 185–200.
Peyriéras, N., Strähle, U., and Rosa, F. (1998). Conversion of zebrafish blastomeres to an endodermal fate by TGF-␤-related signalling.
Curr. Biol. 8, 783–786.
Picker, A., Brennan, C., Reifers, F., Clarke, J.D.W., Holder, N., and
Brand, M. (1999). Requirement for the zebrafish mid-hindbrain
boundary in midbrain polarisation, mapping and confinement of the
retinotectal projection. Development 126, 2967–2978.
Prada, C., Puga, J., Pérez-Méndez, L., López, R., and Ramírez, G.
(1991). Spatial and temporal patterns of neurogenesis in the chick
retina. Eur. J. Neurosci. 3, 559–569.
Raymond, P.A., Barthel, L.K., and Curran, G.A. (1995). Developmental patterning of rod and cone photoreceptors in embryonic
zebrafish. J. Comp. Neurol. 359, 537–550.
Rebagliati, M.R., Toyama, R., Haffter, P., and Dawid, I.B. (1998).
cyclops encodes a nodal-related factor involved in midline signaling.
Proc. Natl. Acad. Sci. USA 95, 9932–9937.
Renucci, A., Lemarchandel, V., and Rosa, F. (1996). An activated
form of type I serine/threonine kinase receptor TARAM-A reveals a
specific signalling pathway involved in fish head organiser formation. Development 122, 3735–3743.
Roessler, E., Belloni, E., Gaudenz, K., Jay, P., Berta, P., Scherer,
S.W., Tsui, L.-C., and Muenke, M. (1996). Mutations in the human
Sonic Hedgehog gene cause holoprosencephaly. Nat. Genet. 14,
357–360.
Rubenstein, J.L.R., and Beachy, P.A. (1998). Patterning of the embryonic forebrain. Curr. Opin. Neurobiol. 8, 18–26.
Li, H.-S., Tierney, C., Wen, L., Wu, J.Y., and Rao, Y. (1997). A single
morphogenetic field gives rise to two retina primordia under the
influence of the prechordal plate. Development 124, 603–615.
Sampath, K., Rubinstein, A.L., Cheng, A.M.S., Liang, J.O., Fekany,
K., Solnica-Krezel, L., Korzh, V., Halpern, M.E., and Wright, C.V.E.
(1998). Induction of the zebrafish ventral brain and floorplate requires cyclops/nodal signalling. Nature 395, 185–189.
Macdonald, R., and Wilson, S.W. (1997). Distribution of Pax6 protein
during eye development suggests discrete roles in proliferative and
differentiated visual cells. Dev. Genes Evol. 206, 363–369.
Sandell, J.H., Martin, S.C., and Heinrich, G. (1994). The development
of GABA immunoreactivity in the retina of the zebrafish (Brachydanio
rerio). J. Comp. Neurol. 345, 596–601.
Macdonald, R., Barth, K.A., Xu, Q., Holder, N., Mikkola, I., and Wilson, S.W. (1995). Midline signalling is required for Pax gene regulation and patterning of the eyes. Development 121, 3267–3278.
Schauerte, H.E., van Eeden, F.J., Fricke, C., Odenthal, J., Strähle,
U., and Haffter, P. (1998). Sonic hedgehog is not required for the
initiation of medial floor plate cells in the zebrafish. Development
125, 2983–2993.
Macdonald, R., Scholes, J., Strähle, U., Brennan, C., Holder, N.,
Brand, M., and Wilson, S.W. (1997). The Pax protein Noi is required
for commissural axon pathway formation in the rostral forebrain.
Development 124, 2397–2408.
Malicki, J., Neuhauss, S.C.F., Schier, A.F., Solnica-Krezel, L., Stemple, D.L., Stainier, D.Y.R., Abdelilah, S., Zwartkruis, F., Rangini, Z.,
and Driever, W. (1996). Mutations affecting development of the zebrafish retina. Development 123, 263–273.
Schier, A.F., and Shen, M.M. (2000). Nodal signalling in vertebrate
development. Nature 403, 385–389.
Schier, A.F., Neuhauss, S.C.F., Helde, K.A., Talbot, W.S., and
Driever, W. (1997). The one-eyed pinhead gene functions in mesoderm and endoderm formation in zebrafish and interacts with no
tail. Development 124, 327–342.
Schmitt, E.A., and Dowling, J.E. (1996). Comparison of topographical
Retinal Neurogenesis in Zebrafish
263
patterns of ganglion and photoreceptor cell differentiation in the
retina of the zebrafish, Danio rerio. J. Comp. Neurol. 371, 222–234.
Shanmugalingam, S., Houart, C., Picker, A., Reifers, F., Macdonald,
R., Barth, K.A., Griffin, K., Brand, M., and Wilson, S.W. (2000). Ace/
Fgf8.1 is required for forebrain commissure formation and patterning
of the telencephalon. Development 127, 2549–2561.
Thisse, C., and Thisse, B. (1999). Antivin, a novel and divergent
member of the TGF␤ superfamily, negatively regulates mesoderm
induction. Development 126, 229–240.
Thisse, C., Thisse, B., Halpern, M.E., and Postlethwait, J.H. (1994).
goosecoid expression in neuroectoderm and mesendoderm is disrupted in zebrafish cyclops gastrulas. Dev. Biol 164, 420–429.
Thisse, B., Wright, C.V.E., and Thisse, C. (2000). Nodal related factors control the antero-posterior patterning of the zebrafish embryo.
Nature 403, 425–428.
Varga, Z.M., Wegner, J., and Westerfield, M. (1999). Anterior movement of ventral diencephalic precursors separates the primordial
eye field in the neural plate and requires cyclops. Development 126,
5533–5546.
Wei, Y., and Allis, C.D. (1998). Pictures in cell biology. Trends Cell
Biol. 8, 266.
Wilson, S.W., Ross, L.S., Parrett, T., and Easter, S.S., Jr. (1990). The
development of a simple scaffold of axon tracts in the brain of the
embryonic zebrafish, Brachydanio rerio. Development 108, 121–145.
Woo, K., and Fraser, S.E. (1995). Order and coherence in the fate
map of the zebrafish nervous system. Development 121, 2595–2609.
Zhang, J., Talbot, W.S., and Schier, A.F. (1998). Positional cloning
identifies zebrafish one-eyed pinhead as a permissive EGF-related
ligand required during gastrulation. Cell 92, 241–251.