Development 122, 735-746 (1996)
Printed in Great Britain © The Company of Biologists Limited 1996
DEV2028
735
Ectopic expression of Hoxa-1 in the zebrafish alters the fate of the mandibular
arch neural crest and phenocopies a retinoic acid-induced phenotype
Daniel Alexandre1,†, Jonathan D. W. Clarke2, Elli Oxtoby3, Yi-Lin Yan4, Trevor Jowett3 and Nigel Holder1,*
1Developmental Biology Research Centre, Randall Institute, King’s College, 26-29 Drury Lane, London WC2B 5RL,
2Department of Anatomy and Developmental Biology, University College, Windeyer Building, Cleveland Street,
UK
London, W1P 6DB, UK
3Department of Biochemistry and Genetics, Medical School, University of Newcastle upon Tyne, NE2 4HH, UK
4Department of Biology, University of Oregon, Eugene, Oregon 97403, USA
*Author for correspondence (email: udbl309@bay.cc.kcl.ac.uk)
†Author’s current address: Laboratoire de Neurogénétique du Développement, Université Montpellier II, C.C. 103, Place Eugène Bataillon, 34095 Montpellier Cedex 5,
France
SUMMARY
Considerable evidence has demonstrated that retinoic acid
influences the formation of the primary body axis in vertebrates and that this may occur through the regulation of
Hox gene expression. In this study, we show that the
phenotype induced by exogenous retinoic acid in the
zebrafish can also be generated by the overexpression of
Hoxa-1 following injection of synthetic RNA into the fertilised egg. The isolation, sequence and expression pattern
of the zebrafish Hoxa-1 gene is described. We show that
exogenously applied retinoic acid causes the ectopic accumulation of Hoxa-1 message during gastrulation in the
hypoblast in the head region. Overexpression of Hoxa-1
following injection of RNA causes abnormal growth of the
anterior hindbrain, duplication of Mauthner neurons in
rhombomere (r) 2 and fate changes of r2 mesenchymal and
neurogenic neural crest. These results are discussed in
terms of the role of Hoxa-1 in controlling anterior
hindbrain patterning and the relationship between
expression of Hoxa-1 and retinoic acid.
INTRODUCTION
regulator of the normal formation of the hindbrain. Overexpression leads predominantly to the conversion of rhombomere
(r) 2 to r4 and results in the alteration in expression of a number
of other regulatory genes, including Hoxb-1 and Hoxa-2. Loss
of function causes loss of r5 and most of r4.
In addition to the Hox genes, the putative developmental signalling molecule, retinoic acid (RA), has also been implicated
in the patterning of the primary body axis in vertebrates. RA
is present during gastrulation in the mouse and the frog (Hogan
et al., 1992; Chen and Solursh, 1994) and RA, exogenously
applied from the early stages of gastrulation, affects patterning
of the anterior-posterior axis in zebrafish (Holder and Hill,
1991; Hill et al., 1995), Xenopus (Durston et al., 1989; Sive et
al., 1990; Papalopulu et al., 1991) and mammals (Morriss,
1972; Morriss-Kay et al., 1991; Kessel and Gruss, 1991;
Conlon and Rossant, 1992; Marshall et al., 1992; Simeone et
al., 1995). In such experiments, one consistent effect of RA is
to alter the formation of the hindbrain, where the segmental
rhombomeres are to some extent posteriorised (Papalopulu et
al., 1991; Marshall et al., 1992; Hill et al., 1995). Exogenously
applied RA leads to the alteration of expression of a number
of regulatory genes in the hindbrain (Morriss-Kay et al., 1991;
Conlon and Rossant, 1992; Marshall et al., 1992) but it remains
unclear how the phenotype is generated. A number of lines of
evidence indicate that Hoxa-1 is a potential target for RA. As
Genetic studies, principally in Drosophila and in mouse, have
shown that Hox genes are key regulators of positional information during development (see reviews by McGinnis and
Krumlauf, 1992; Krumlauf, 1994). Loss-of-function alleles in
the mouse, for example, generate either homeotic transformations or abnormal development of the region of the embryo
associated with the anterior limit of normal expression of the
gene (eg Mark et al., 1993; Rijli et al., 1993; Ramirez-Solis et
al., 1993). Homeotic changes are also a consequence of overexpression of a number of different members of the vertebrate
Hox clusters (eg Kessel et al., 1990; Lufkin et al., 1992; Zhang
et al., 1994). The homeotic changes primarily affect the
segmental organisation of the primary body axis such as the
formation of inappropriate vertebrae at specific levels (Kessel
et al., 1990; Jegalian and DeRobertis, 1992; Ramirez-Solis et
al., 1993) or alterations to specific skull bones (Lufkin et al.,
1992; Rijli et al., 1993; Gendron-Maguire et al., 1993). It is
clear, therefore, that the Hox genes are important in the establishment of the primary body axis. This is best illustrated by
the mouse Hoxa-1 gene which has been studied both by overexpression (Zhang et al., 1994) and by loss-of-function
mutation (Lufkin et al., 1991; Chisaka et al., 1992; Mark et al.,
1993; Carpenter et al., 1993) and has been shown to be a key
Key words: Hoxa-1, neural crest, pharyngeal arches, retinoic acid,
homeosis
736
D. Alexandre and others
the most 3′ in the Hoxa cluster, it is most sensitive to induction
by RA in NT2 cells (Simeone et al., 1990) and an RA response
element has been identified in the Hoxa-1 gene that is
necessary for RA induction (Langston and Gudas, 1992). It is
also expressed from the onset of gastrulation (Murphy and Hill,
1991), which, as mentioned above, is the period in the
embryonic process when formation of the primary body axis
is most sensitive to exogenous RA treatment The link between
the expression and function of Hoxa-1 and the endogenous role
of RA is further strengthened by similarities in aspects of the
phenotypes generated by either Hoxa-1 overexpression or
exogenous supply of RA to the embryo. This has been shown
in the mouse by Marshall et al. (1992) who demonstrated that
RA causes a transformation of r2 into r4, a result that is
mirrored in mouse embryos ectopically expressing Hoxa-1
(Zhang et al., 1994). Similarly, RA and ectopic Hoxa-1 both
lead to altered expression of Hoxb-1 and Hoxa-2 in the
hindbrain (Conlon and Rossant, 1992; Kessel, 1993; Zhang et
al., 1994).
In previous studies, we have characterised the effect of
exogenous RA on the development of the zebrafish hindbrain
(Holder and Hill, 1991; Hill et al., 1995). Following the application of a relatively low concentration of RA, a precise
phenotype is generated in the anterior hindbrain. The limits of
the effect are the caudal midbrain rostrally and the r4/5 border
caudally and include, in some respects comparable to the
mouse (Marshall et al., 1992), the partial transformation of r2
to r4. For the reasons outlined above, we wished to analyse the
expression and function of Hoxa-1 in the zebrafish in order to
establish whether it is the likely target gene for exogenously
applied RA and whether alteration in its expression is partly or
wholly responsible for the characteristic RA-induced hindbrain
phenotype. To this end, we have isolated and characterised the
zebrafish Hoxa-1 cDNA and show that its expression pattern
is essentially that seen in the mouse. The expression of the gene
is affected by exogenously applied RA from the onset of gastrulation. We analyse the function of Hoxa-1 by RNA injection
into the fertilised egg and characterise a reliably generated
phenotype that includes alterations to the anterior rhombomeres and to the anterior hindbrain neural crest.
The neural crest in the head can be divided into two broad
populations, the neurogenic precursors, which will differentiate into neurons associated with the peripheral head ganglia,
and the mesenchymal precursors, which migrate to the pharyngeal arches and differentiate into Schwann cells, pigment
cells and cartilage and connective tissue cells (review by Le
Douarin et al., 1994). In the head of the zebrafish, the neurogenic neural crest is a region of cells located most laterally
in the dorsal crest primordium (Schilling and Kimmel, 1994)
and these authors have shown that neural lineage restriction
occurs prior to migration. The mesenchymal crest lies more
medially in the premigratory mass and is also restricted in
terms of eventual fate prior to migration. The phenotype that
we describe, which has not been observed previously in comparable Hoxa-1 overexpression experiments (Zhang et al.,
1994), is an alteration of the pharyngeal skeleton whereby a
derivative of the first arch, the mandible, is not formed and
the second arch-derived hyoid cartilage is enlarged. This
alteration is linked to abnormal migration of the mandibular
arch crest. This is coupled with a fusion of cranial ganglia V
and VII. The phenotype generated by Hoxa-1 overexpression
in zebrafish is very similar to that produced by exposure to
RA.
MATERIALS AND METHODS
(A) Cloning of the zebrafish Hoxa-1 gene
A polymerase chain reaction was performed on zebrafish genomic
DNA using primers based on the 5′ and 3′ ends of the labial-like
homeobox of murine Hoxa-1 and Hoxb-1. The 5′ primer was
CTGCAGCGCACCAACTTCACCACNAA(A/G)CA (lab+) and the 3′
primer was CTCGAGCTCGCGCTCGCGCTTCTT(T/C)TG(T/C)TT
(lab−). 0.5 µg of zebrafish genomic DNA was amplified in 100 µl
reaction containing 50 mM KCl, 10 mM Tris-HCl pH 9.0 at 25°C,
1.5 mM MgCl2, 0.01% gelatin, 0.1% Triton X-100, 0.2 mM each
dNTPs, 0.5 µg of each primer and 2.5 units of Taq polymerase
(Promega). After an initial denaturation of 5 minutes at 94°C, thirty
rounds of amplification were performed, a single cycle being 2
minutes at 55°C, 3 minutes at 72°C and 2 minutes at 94°C. A final
extension reaction was performed for 10 minutes at 72°C. The gel
purified products were treated with polynucleotide kinase and then
Klenow before blunt-end ligating into the SmaI site of Bluescript
KS−. Seven plasmids were identified as having the same homeobox
sequences and these sequences were used to design primers for performing 5′ and 3′ RACE (Frohman and Martin, 1989).
For 3′ RACE, the initial reverse transcription was performed in 50
mM Tris-HCl, pH 8.15 at 41°C, 6 mM MgCl2, 40 mM KCl, 1 mM
DTT containing 1 mM of each dNTP, 10 units of RNasin, 1 µg of
zebrafish poly(A)+ RNA and 2.5 pmole of the 57-mer (dT)17 -R1-R0
primer (Frohman and Martin, 1989) with 10 units of avian myeloblastosis virus reverse transcriptase (AMV-RT) for 2 hours. The reaction
mixture was diluted to 1 ml with TE (10 mM Tris-HCl pH 8.0, 1 mM
EDTA). For first round amplification, 5 µl of the diluted cDNA pool
was mixed with 25 pmol each of a gene-specific primer (3H29A in
Fig. 1) and R0 (Frohman and Martin, 1989) in a 50 µl PCR cocktail
with the same ingredients as in the original genomic DNA reaction.
After an initial denaturation step of 95°C for 7 minutes, the reaction
was cooled to 70°C and 2.5 units of Taq polymerase added, then thirty
cycles of 94°C for 2 minutes, 50°C for 2 minutes and 72°C for 3
minutes performed with a final extension of 15 minutes at 72°C. The
reaction was diluted 1:20 with TE and a second amplification
performed with 1 µl of the diluted DNA and 25 pmol each of a second
nested gene-specific primer (3H29B in Fig. 1) and R1 (Frohman and
Martin, 1989). The conditions for the amplification were otherwise
the same as for the first round above. A single fragment was amplified
and this was gel purified, treated with kinase and Klenow before subcloning into the SmaI site of Bluescript KS−. Before subcloning the
amplified fragment, we confirmed that it was as expected by making
a 10−6 dilution of the purified fragment and showing that it could be
reamplified with the 3H29C and R1 primers.
For 5′ RACE, the initial reverse transcription was the same as for
the 3′ reaction except it incorporated 1 pmole of a gene-specific
primer (5H29A in Fig. 1) instead of the 57-mer. The cDNA reaction
was diluted to 1 ml with TE and the excess primer was removed on
a Centricon 100 (Amicon Corp.) ultrafiltration spun column. The final
retenate was concentrated to 10 µl before dATP tailing with terminal
deoxynucleotide transferase (BRL) and then diluting to 500 µl with
TE. The first round amplification involved 5 µl of diluted tailed cDNA
and 25 pmoles each of 5H29A and R0 with an additional 2 pmol of
the 57-mer in a 50 µl reaction volume. Denaturation followed by thirty
cycles was as for 3′ RACE. Second round amplification was with 1
µl of a 1:20 dilution of the first reaction and 25 pmol each of a second
nested gene-specific primer (5H29B in Fig. 1) and R1. Amplification
conditions were as for the previous reaction. The final reaction yielded
three major amplification bands. Before subcloning the amplified
fragments, we confirmed that they were derived from the same gene
Ectopic expression of Hoxa-1
by making a 10−6 dilution of the purified fragments and showing that
they were reamplified with the 5H29C and R1 primers. All three were
purified and subcloned as previously described and on sequencing
they were shown to be subsets of the same reverse transcription
reaction. It would appear that the reverse transcriptase quite frequently
fell off the RNA template prematurely thereby generating two smaller
cDNAs.
Primers H16(72) and H16(833), in Fig. 1, were used to PCR amplifications of the 3′RACE cDNA pool and also from genomic DNA.
Conditions for amplification were the
same as for that with the lab+ and lab−
primers. The amplified product derived
from the genomic DNA was slightly
larger than the cDNA product.
Sequencing of the two amplified
fragments revealed a single 89 base pair
intron in the genomic sequence. The
cDNA amplification product was used
as a probe to a zebrafish genomic
library in lambda DASH® II (Stratagene). A single clone was selected that
had the region of homology in the
middle of the insert. The insert was
subcloned into Bluescript SK− and
restriction mapped. The region of
homology with the cDNA clone was
within a single 1335 bp SinI fragment
(Fig. 1). This SinI fragment was gel
purified and, after addition of BamHI
linkers, was subcloned into the BglII
site of the transcription vector pSP64T
(Krieg and Melton, 1984). The intron
was removed by excising the XcmIBssHII fragment (Fig. 1) and replacing
it with the equivalent fragment taken
from the cDNA clone produced by
amplification with primers H16(72) and
H16(833). The resulting plasmid,
zebrafish
Hoxa-1
intronless
(SinI)BamHI/pSP64T, was used to
generate
full-length,
capped,
polyadenylated transcripts with SP6
RNA polymerase. A negative control
for injections was made by taking
the
zebrafish
Hoxa-1-intronless
(SinI)BamHI/pPSP64T plasmid, linearising with XcmI, blunt-ending the
fragment with T4 DNA polymerase and
recircularising with T4 ligase. This
changes the open reading frame such
that the SP6 RNA polymerase transcript
contains a stop codon just into the
homeobox region and thus will produce
a truncated Hoxa-1 polypeptide that
lacks the homeodomain.
(B) Injection of mouse and fish
Hoxa-1 RNA
The mouse Hoxa-1 cDNA p480
encoding the 331 amino acid protein
was excised with EcoRI and PstI and
subcloned initially into Bluescript KS−.
BamHI was used to transfer the gene
into the BglII site of the pSP64T vector
and the orientation was checked by
restriction mapping. A negative control
construct was produced by removing
737
most of the homeobox by cutting with HincII and BglII. The ends
were then blunt ended with mung bean nuclease and ligated producing
an in frame fusion which was checked by sequencing.
RNA synthesis, purification and injection into fertilised eggs of 300
to 500 pl of RNA at a concentration of 100 µg/ml was carried out as
described by Griffin et al. (1995). Prior to injection, all RNAs were
tested by in vitro translation (Boehringer Mannheim) according to
manufacturers recommendations. Translation reactions were
separated by SDS-PAGE and visualised by autoradiography.
Fig. 1. Genomic DNA sequence of the zebrafish Hoxa-1 gene. This is derived from both genomic
clones and cDNAs as described in the Materials and Methods. The positions of the primers used
for the polymerase chain reactions used to amplify the homeobox region (lab+ and lab−), for 3′
RACE (3H29-A, 3H29-B, 3H29-C) and 5′RACE (5H29-A, 5H29-B, 5H29-C). The 89 base pair
intron and the homeobox regions are underlined. Putative translations of the two exons, the intron
and primer locations are shown below the corresponding DNA sequence. The 3′ end of the
genomic sequence is the polyadenylation site for the transcript. The restriction sites are those used
in subcloning and in making the full-length cDNA construct for in vitro transcription.
738
D. Alexandre and others
(C) Maintenance of fish and RA treatment
Breeding fish were maintained at 28.5°C on a 14 hour light/10 hour
dark cycle. Embryos were collected by natural spawning and raised
in aquarium water at 28.5°C and staged according to Westerfield
(1995). RA treatment was also performed according to previously
described protocols (Holder and Hill, 1991); in the current study, RA
was given at a concentration of 10−7 M. All exposures were at 50%
epiboly, the onset of gastrulation (Westerfield, 1995) for 1 hour at
28.5°C.
(D) Antibody and cartilage staining
3A10, an antibody originally raised in a screen for rat floor plate tissue
by Dr Jane Dodd, recognises exclusively the Mauthner neuron in the
zebrafish at the stages used in this study (Hatta, 1992). The anti-acetylated tubulin and 4D9 antibodies were purchased commercially. The
staining procedures were those described previously (Holder and Hill,
1991). For Alcian blue staining, 4 day old larvae were fixed in 4%
paraformaldehyde, rinsed in PBS and stained for 2 hours in 0.01%
Alcian blue in 70% ethanol/30% glacial acetic acid. They were then
destained in the acidic alcohol, rinsed once with 40% ethanol in PBS
then in PBS alone and cleared in 0.5% potassium hydroxide for a few
minutes. Specimens were then rinsed and mounted in glycerol.
(E) Whole-mount in situ hybridisation
For single colour in situ mRNA probes to zebrafish krx-20 (Oxtoby
and Jowett, 1993), pax 2 (Krauss et al., 1991), dlx 2 (Akimenko et al.,
1994) and Hoxa-1 were prepared and used as described by Oxtoby
and Jowett (1993). For two colour in situ hybridisation, krx-20 was
labelled with fluorescein and dlx-2 with DIG and detected according
to the methods described by Jowett and Lettice (1994) as modified by
Hauptmann and Gerster (1994) except that glycine treatment was
extended to 40 minutes.
(F) Retrograde labelling of reticulospinal neurons
4 day embryos were immobilised in 1% low melting point agarose in
PBS. The spinal cord was completely transected at the level of the
hindgut with a sharpened tungsten needle. The dye (lysinated
rhodamine dextran, Molecular Probes D-1817) was then applied as a
thick semi-solid with another sharpened tungsten needle to the rostral
stump of the cut spinal cord. The embryo was then released into a
solution of 10% Hank’s saline for 2 hours to allow transportation of
the dye. The embryo was then fixed in 4% paraformaldehyde in PBS
overnight at 4°C.
The brains from labelled embryos were then carefully dissected free
of surrounding tissue and mounted ventral side uppermost in 90%
glycerol/10% PBS on a glass microscope slide, under a glass coverslip
supported by four spots of silicon gel. Permanent mounts were painted
around the edge of the coverslip with nail varnish to prevent evaporation.
Confocal images were viewed on a Leica confocal microscope
using standard Leica software.
RESULTS
Cloning of the zebrafish homologue of murine
Hoxa-1
We initially used primers complementary to the conserved
ends of the labial-like homeobox of the murine Hoxa-1 and
Hoxb-1 genes to perform the polymerase chain reaction on
zebrafish genomic DNA. They amplified a 173 bp DNA
fragment which when subcloned and sequenced proved to be
a DNA species with the homeobox sequence shown in Fig. 1.
Although the primers would have been expected to produce
two different labial-like sequences corresponding to Hoxa-1
and Hoxb-1, we only obtained the one type of sequence in 10
different clones. We then designed sets of nested primers based
on the sequence of the zebrafish homeobox for use in 5′ and
3′ Rapid Amplification of cDNA Ends, RACE (Frohman and
Martin, 1989). The starting material for the reactions was
poly(A)+ RNA from 12-24 hour old zebrafish embryos. The 3′
RACE produced a single amplified DNA species with the
sequence from 3H29B to the 3′ end of the sequence in Fig. 1.
The 5′ RACE amplification gave three differently sized
fragments which proved to be derivatives of a single 742 bp
fragment with the sequence from the 5H29B primer to base
137 in Fig. 1. (Note that these clones also lacked the intron
sequence shown in Fig. 1).
We concluded that both the cDNA sequences were derived
from the same transcript for the following reasons. They
hybridised to identically sized transcripts in northern blots of
the original poly(A)+ RNA population (not shown). They also
gave similar patterns of hybridisation in Southern blots of
genomic DNA cut with different restriction enzymes (not
shown). We designed primers from opposite ends of each
cDNA (H16(72) and H16(833)) and used them to amplify
DNA from a reverse transcribed RNA pool and they gave a
single DNA species with a sequence that was predicted by
joining our original two overlapping cDNAs.
Amplification of genomic DNA using the same two primers
gave a slightly larger than predicted fragment, which on
sequencing proved to have an 89 bp intron located at the valine
codon upstream from the start of the homeobox (Fig. 1). The
position of this intron corresponds to the location of the introns
found in Hoxa-1 (LaRosa and Gudas, 1988), Hoxb-1 (Frohman
et al., 1990) and Hoxd-1 (Frohman and Martin, 1992a) and the
chick Hoxb-1 gene (Sundin et al., 1990; Sundin and Eichele,
1990) just upstream of the homeobox.
From the translation of the open reading frame harbouring
the homeobox, it appeared that the largest of the 5′RACE
clones lacked the start of the protein encoding open reading
frame which would be essential for making full-length transcripts for injection into embryos. We therefore screened a
zebrafish genomic DNA library with the cDNA clone and
obtained a lambda clone which harboured the region of
homology in the middle of the insert. The insert was subcloned
into Bluescript SK− and restriction mapped. Analysis of this
clone revealed that the transcribed region was within a 1335
base pair SinI fragment. This was subcloned into the in vitro
transcription vector pSP64T and the intron was removed by
replacing the XcmI-BssHII fragment, which harboured the
intron with the equivalent fragment from the cDNA clone. This
plasmid, Hoxa-1 intronless (SinI)BamHI/pSP64T, is suitable
for generating full-length polyadenylated transcripts for
injection into embryos.
A comparison of the putative zebrafish polypeptide shows
that the protein most closely resembles that encoded by the
Hoxa-1 gene (Fig. 2). There is considerable homology within
the homeobox and also around the conserved hexapeptide
region (WMKVKR) characteristic of many other members of
the Hox family (Duboule et al., 1988). The sequence comparison alone did not allow us to unambiguously assign the
zebrafish gene as being homologous to Hoxa-1.
The zebrafish Hoxa-1 gene maps to the middle of linkage
group 12 (Gates and Postlethwait, unpublished observation see Postlethwait et al., 1994).
Ectopic expression of Hoxa-1
Expression pattern of Hoxa-1 in whole mounts
To confirm our designation of the zebrafish gene as the
homolgue of the murine Hoxa-1 and not Hoxb-1 or Hoxd-1,
we performed in situ hybridisations on embryo whole mounts.
The original probe used was a mixture of two digoxigeninlabelled antisense RNAs made from the RACE cDNA
subclones into Bluescript. When used separately these probes
both gave the same pattern of signals. (Both probes contain
part of the homeobox region; under our conditions of hybridisation, this did not cause any problems of cross-hybridisation
with other homeobox-containing transcripts.)
The first appearance of Hoxa-1 is at 50% epiboly as the germ
ring forms (Fig. 3A,B). It produces a semicircle of stain which
is broken in the middle and is located just above the margin.
It is centred around the localised thickening of the germ ring
known as the shield, at the dorsal surface of the gastrula. As
the cells migrate towards the vegetal pole over the surface of
the yolk and begin to involute, the domain of expression
broadens. The upper boundary nearest the animal pole remains
relatively sharp and the gap in the region of the shield begins
to close (Fig. 3C,D). At the tailbud stage, the Hoxa-1 domain
is quite broad and indistinct but has a clear anterior border (Fig.
3E-H). As cells converge towards the dorsal surface, a stripe
appears at the midline. This is probably caused by the concentration of cells as they gather along the embryonic A/P axis
rather than an increase in expression. As somitogenesis progresses, the Hoxa-1 signal is present in the spinal cord and in
the endoderm (Fig. 3I-J). Endodermal expression has a more
anterior boundary than that in the spinal cord. This expression
pattern is consistent with the gene being Hoxa-1 rather than
Hoxb-1 since the latter gene in other vertebrates shows a
discrete domain of expression in rhombomere 4 (this designation assumes an expression pattern for fish labial-like Hox
genes that is similar to that found in other vertebrates).
Similarly it is not likely to be the homologue of Hoxd-1
because this gene is expressed only in the mesoderm (Frohman
and Martin, 1992b).
In order to assess the anterior border of expression of Hoxa1 during the neural plate stages, double in situs with Hoxa-1
and krx-20 were performed between 10 and 16 hours (Fig. 3KO). At the time that the anterior (r3) stripe of krx-20 first appears
there is a gap between it and the anterior boundary of Hoxa-1.
The second krx-20 (r5) stripe then appears at the boundary of
Hoxa-1, although some Hoxa-1-expressing cells are also
present in the lateral regions of r4 (see Fig. 3M). As the cells
converge on the dorsal midline, the rostral boundary of Hoxa1 recedes caudally to approximately the boundary between the
hindbrain and spinal cord. These results indicate that the most
anterior extent of Hoxa-1 is probably within r4 or at the r3/r4
boundary. It is difficult to be certain of this because the Hoxa1 expression domain may already be receding by the time that
krx-20 is initially expressed and the expression of Hoxa-1 in r4
is only clear in the lateral regions (Fig. 3M).
The effect of RA on expression of Hoxa-1
If the effect of RA on hindbrain development is linked to Hoxa1, it is assumed that RA will alter the expression of the gene.
To see if this is the case, embryos were treated with RA at the
onset of gastrulation in a manner that is known subsequently
to cause abnormal development of the anterior hindbrain
(Holder and Hill, 1991). Treated and control embryos (treated
739
with DMSO only) were then fixed at times through gastrulation, tail bud and early somite stages and hybridised with
Hoxa-1 probe under identical conditions. The results, shown
in Fig. 4, indicate an upregulation of Hoxa-1 within an hour of
treatment as compared with controls. By 80% epiboly (Fig.
4A), in comparison to an identically treated control, Hoxa-1 is
significantly upregulated around the germ ring and expresses
in the axial midline in the hypoblast and not in the epiblast
(Fig. 4A-F). The increased level of expression in the hypoblast
maintained through tail bud and early somite stages when the
ectopic expression can be seen to be in the prechordal
mesoderm (Fig. 4C-F). In addition, it appears with respect to
the midbrain pax 2 stripe that the expression of Hoxa-1 in the
ectoderm is not ectopic, the boundary being at the same
location as in controls (Fig. 4B,C).
Assessing the function of Hoxa-1 by injection of
RNA
RNA encoding the mouse or zebrafish Hoxa-1 protein was
injected into the fertilised egg or in both blastomeres of the 2cell embryo. The majority of the analysis was performed on
embryos injected with the mouse wild-type and control RNA,
which was lacking the homeobox. In addition, the zebrafish
Hoxa-1 RNA was used for the analysis of crest derivatives, the
visceral skeleton and cranial ganglia. In general, the survival
rates of injected embryos with either wild-type or control RNA
was close to 100% at 24 hours. After this time, the mortality
of embryos injected with the wild-type RNA increases to reach
about 30% by 5 days. In contrast, the embryos injected with
control RNA show no increased mortality compared to uninjected fish.
In order to assess the degree of mosaic distribution of the
injected RNA, in situ hybridisation was performed using a
probe to the transcribed flanking globin sequences which are
generated by the pSP64T vector used to make injected
zebrafish Hoxa-1. Following injection into the single cell egg,
by the 32-/64-cell stage, the RNA is present in groups of cells
which vary in size from 10 to 50% of the total blastomeres with
the majority of embryos showing a quarter of blastomeres
positive for the RNA (Fig. 5). At this stage, the groups of cells
tended to be together rather than mixed with cells not containing the RNA. As assessed by in situ hybridisation, this
ectopic expression persists at least to tail bud stages when
expressing cells are present in the derivatives of the epiblast
and hypoblast. Distribution of expression over the embryo at
this stage is extensive and random with expressing cells present
in all regions of the embryo (data not shown).
The phenotype of embryos injected with Hoxa-1
RNA
For the overall analysis of function by overexpression, a total
of not less than two thousand eggs were injected with Hoxa-1
RNA. For a typical experiment, injections were performed on
about 200 embryos and were repeated at least once. The
phenotype described below was never observed in embryos
injected with a control RNA. For illustrative purposes, the
specimens were chosen to show the most complete phenotype,
which occurred in approximately 20% of injected embryos.
Due to mosaicism, in the remaining injected embryos, the phenotypes were generally restricted to one side or were present
dorsally and not ventrally.
Fig. 3. Expression pattern of Hoxa-1in whole-mount in situ hybridisations. (A,B) 50%
epiboly. A is the view from the animal pole dorsal side towards the bottom, B is the same
embryo as in A but tilted to show more of the dorsal side. (C,D) About 70% and 80% epiboly
respectively, viewed from the dorso-vegetal side, note the broadening of the domain of
Fig. 2. Alignment of the protein sequences of Hoxa-1 and Hoxb-1 homologues. Alignment
was by the Clustal method using the Megalign™ programme from DNASTAR™.
expression towards the vegetal pole and also the closing of the gap at the dorsal
midline. (E,F) 100% epiboly. E is a dorsal view and F is the same embryo
viewed form the side, dorsal to the right and rostral end of the embryo is at the
top. (G,H) Progressively older embryos after the tailbud stage viewed from the
dorsal side. Note the sharp anterior boundary of the expression domain and the
increase in signal at the dorsal midline where the cells are converging to form
the spinal cord. (I,J) 20 hour old embryo showing the strong signals from the
spinal cord, the endoderm (arrowheads), the tailbud, the regions perpendicular
to the A/P body axis at the level of the otic vesicle. (K-O) Embryos in a
sequence from 10 to 16 hours showing the expression of Hoxa-1 and krx-20.
The boundary of Hoxa-1 at the time krx-20 first expresses is just caudal to r3.
By 16 hours, this rostral Hoxa-1 boundary has receded to the spinal
cord/hindbrain border. Expression can be seen in the lateral regions of r4 prior
to the signal receding – arrowed in M. Scale bar, 100 µm.
740
D. Alexandre and others
Ectopic expression of Hoxa-1
741
Fig. 4. Retinoic acid causes the ectopic expression of
Hoxa-1. (A-C) Pairs of embryos that have hybridised to
probes to Hoxa-1 and pax 2 under indentical
conditions. The embryos on the right of each pair have
been treated with RA on the onset of gastrulation. At
this dose of RA, Hoxa-1 is ectopically expressed by
70% epiboly (A) in the axial midline (black arrow) and
around the germ ring as seen from the animal pole. By
the 1-somite stage (B - tilted animal pole view) pax 2
(white arrow) is first expressed in the prospective
midbrain and, by this stage, the rostral limit of the
Hoxa-1 signal in the treated embryos is no further
anterior than in the control. A small patch of
ectopically expressed Hoxa-1 lies anteriorly (black
arrow). (C) The same embryo as seen in B but viewed
laterally with anterior to the top and dorsal to the right.
The white arrow indicates the normal pax 2 expression
and the black arrow the ectopic Hoxa-1 expression
which is in the prechordal mesoderm. (D-F) A series of
high power views of optical sections through the
midline of treated embryos showing ectopic expression
of Hoxa-1 only in the hypoblast. (D) Ectopic
expression is evident at the late shield stage.
(E) Ectopically expressing cells are seem to be in the
hypoblast. (F) The same embryo as that shown in B and C. Ectopic expression is present in the prechordal mesoderm. The arrow indicates
the anterior extent of the normal expression of Hoxa-1 in the ectoderm. pcm, prechordal mesoderm; h, hypoblast; e, ectoderm. Scale bar,
100 µm.
(1) General appearance
Externally the injected embryos developed normally through
gastrulation and subsequently formed the main parts of the
body axis. Under the dissecting microscope, approximately
70% of the injected embryos showed abnormal development
of the anterior hindbrain in that the neural tube in this region
appeared twisted to one side or expanded (Fig. 6 and see
below). The remaining CNS, anterior notochord and hatching
gland were normal. The only other obvious abnormality
became evident after several days of development at the most
anterior end of the body when the lower jaw was abnormally
formed in 77% of embryos that survived to 5 days (Fig. 8F).
expression in r5 was invariably normal (Fig. 6B). The region
of r3 was distorted and appeared larger than normal. Morphologically the midbrain and forebrain appeared normal. In one
batch of injected embryos that showed an abnormal phenotype
in the anterior hindbrain, the antibody 4D9 was used to assess
the appearance of the engrailed-expressing region of the caudal
midbrain. In all embryos, 4D9-expressing cells were present in
a pattern that appeared essentially normal (data not shown)
suggesting the normal development of this region of the brain.
The reticulospinal complex offers an excellent assay for the
normal development of the hindbrain because many of the cells
of this complex are individually identifiable and they are
arranged in a segmental manner that reflects their rhombomeric
(2) Development of the hindbrain
This was assessed in injected embryos and controls using a
probe for zebrafish krx-20, the antibody 3A10 to locate the
Mauthner neurons and by analysing the pattern of neurons in
the reticulospinal complex by neuronal tracing using lysinated
rhodamine dextran. krx-20 is normally strongly expressed in
r3 and r5 at 20 hours of development (Oxtoby and Jowett,
1993); therefore, injected and control embryos were analysed
at this time. Control embryos invariably showed the normal
pattern of two stripes (Fig. 6A). In contrast, the experimental
embryos showed a dramatic alteration in expression in r3 but
Fig. 5. Embryo at the 132cell stage hybridised with a
probe to pSP64T vector
sequences following
injection of Hoxa-1 synthetic
RNA into the fertilised egg.
The RNA distributes to
approximately one third of
the blastomeres in this case.
Scale bar, 100 µm.
Fig. 6. Photomicrographs of the anterior hindbrain region of 20 h
embryos hybridised with a probe for krx-20. (A) Control showing
krx-20 expressing in r3 and r5. (B,C) Examples of embryos injected
with mouse Hoxa-1 RNA at the single cell stage. r5 is normal in both
cases but r3 expression of krx-20 is abnormal. The anterior hindbrain
appears distorted; thus r3 is split dorsally in B and is laterally
displaced in C. Ectopic regions of krx-20 expression, such as that
arrowed most anteriorly in C, were occasionally seen. Scale bar,
100 µm.
742
D. Alexandre and others
origin (Kimmel et al., 1982, 1985; Metcalfe et al., 1986; Hill
et al., 1995). The Mauthner cell is one of the individually identifiable neurons and is characteristic of r4. It is conveniently
identified in whole-mounted embryos by the 3A10 antibody at
32 hours of development (Hatta, 1992; Hill et al., 1995).
Analysis of control and Hoxa-1-injected embryos at this time
showed the Mauthner neuron to be duplicated at more anterior
hindbrain levels (Fig. 7A,C). Without additional markers, it is
difficult to be sure of the rhombomeric identity of the duplicated Mauthner cell; therefore, we assessed the reticulospinal
complex by axonal retrograde labelling. This procedure was
carried out at 4 days of development on control and experimental embryos. In approximately 30% of successful preparations, duplication of the Mauthner cell was evident in r2 in
Hoxa-1-injected embryos but never in controls (Fig. 7D,E);
this occurred either on one (Fig. 7B) or both sides (Fig. 7E).
Duplicated Mauthner neurons were occasionally found in r4
and, in a single case, a Mauthner cell was duplicated in r3. The
Mauthner cell was the only identified neuron that we could be
certain was duplicated and it is certain that when duplicated in
r2 other identified cells in this rhombomere, such as the Rol 2
cell, were of r2 character, making it a hybrid structure.
(3) Neural crest-derived structures
The derivatives of the mesenchymal crest are altered in the
experimental embryos. This is shown with respect to the developing pharyngeal skeleton, which is derived from the crest cells
in each pharyngeal arch. It is the abnormal development of the
jaw apparatus that gives the front of the larvae a truncated
appearance (Fig. 8F). To assess the phenotype, the normal
ventral or viscerocranium and dorsal head skeleton or neurocranium was revealed by staining the cartilage with Alcian blue
in 4 day old control embryos (Fig. 8A,C). In the viscerocranium
(Fig. 8A), the first (mandibular) arch neural crest gives rise to
the Meckel’s cartilage (mandible) and the palatoquadrate
(maxilla). These first arch structures form a characteristic high
arch appearance and lie anterior to the shorter and thicker
second (hyoid) arch-derived ceratohyal and hyosymplectic cartilages. More posteriorly lie a series of five branchial arch cartilages associated with the gills. The neurocranium (Fig. 8C)
differentiates at this time and consists of the posterior and
mesoderm-derived parachordals, the anterior limits of which lie
parallel to the anterior extent of the notochord and the more
anterior trabeculae, which are neural crest-derived (Langille and
Hall, 1987, 1988). At their anterior extent, the trabeculae fuse
and expand to form the ethmoid plate (Fig. 8C). In experimental embryos (Fig. 8B), the cartilagenous derivatives of the first
arch (Meckel’s cartilage and the palatoquadrate) are absent;
however, the second arch-derived ceratohyal cartilages are
formed in their normal location but are thicker and partially
duplicated medially. The hyosymplectic and the 5 branchial
arch cartilages were generally normal but were occasionally
under developed. The neural crest-derived component of the
neurocranium is also malformed; the trabeculae are not
connected to the most anteriorly positioned ethmoid plates
which can be clearly identified by their proximity to the
olfactory pits. The trabeculae (arrowed in Fig. 8B) instead curve
ventrally as they converge rostrally between the eyes (Fig. 8D)
and come downwards to the level of the ceratohyals.
Having established that Hoxa-1 overexpression alters the
anterior head skeleton, it was necessary to establish if RA has
the same effect. Alcian blue preparations were made of
embryos treated with RA at the onset of gastrulation as before.
At 4 days, it is clear that RA causes a distinct and abnormal
differentiation of the Meckel’s cartilage which is much reduced
or absent. The palatoquadrate and the ceratohyals are fused
(Fig. 8E) making this phenotype very similar to the jaws that
develop following Hoxa-1 overexpression.
The derivatives of the neurogenic crest were assessed by
anti-acetylated tubulin antibody staining in injected and control
embryos at 24 hours. At this time, the cranial ganglia form
identifiable populations of cells. The trigeminal (Vth) ganglion
lies behind the eye, the facial (VIIth) ganglion immediately
anterior and at the border of the otic vesicle respectively, and
the posterior lateral line ganglion lying postotically (Fig. 9A).
In wild-type embryos, the Vth ganglion crest component
comes from the most anterior hindbrain stream whereas the
VIIth ganglion is generated by the stream opposite r4. In over
80% of embryos injected with wild-type zebrafish Hoxa-1
RNA, the Vth and VIIth ganglia are fused forming a large
structure just anterior to the otic vesicle (Fig. 9B).
In order to assess whether this abnormal phenotype reflected
an altered pattern of migration, we performed in situs with a
probe for dlx2. Dlx2 has been shown by Akimenko et al. (1994)
to be a good marker for migrating hindbrain neural crest. In
control embryos with between 10 and 20 somites, three streams
of crest are visible in the hindbrain region (Fig. 10A). Cells in
the most anterior of these enter the mandibular arch and cells
in the preotic (middle) stream enter the hyoid arch (Akimenko
et al., 1994). In Hoxa-1-injected embryos, only two streams are
evident, the anteriormost is absent and the stream anterior to
the otic placode is larger than normal (Fig. 10B); the postotic
stream is unaffected in experimental embryos. The fused
preotic stream lies opposite r4 and stretches round the aberrantly formed r3 as seen in embryos in which double in situs
have been performed with krx-20 and dlx-2. (Fig. 10C).
DISCUSSION
Hoxa-1 overexpression following RNA injection into the fertilised zebrafish egg results in a defined series of abnormalities
in the embryo focused on the developing anterior region of the
hindbrain and associated neural crest-derived structures. The
localised region of the embryo that is affected is despite the
wide distribution of the RNA throughout the embryo in cells
derived from 25-50% of blastomeres. Localised effects on the
anterior hindbrain suggest that the abnormal expression pattern
is altering development at the anterior extent of the normal
expression domain, a result that is consistent with studies with
other Hox genes, such as Hoxa-2 (Rijli et al., 1993) and Hoxb4 (Ramirez-Solis et al., 1993). The localised effect of altering
Hoxa-1 expression is consistent with the results of overexpression and loss-of-function experiments with this gene in the
mouse. The phenotypes that are seen are best understood in
terms of the normal expression pattern of the gene. In the
mouse, the anterior extent of expression of Hoxa-1 is the r3/4
boundary. The double labels with krx-20 show that this is likely
to be the case also for the fish although it is only clear in the
lateral regions of r4. However, this expression is transient and
the gene is subsequently down-regulated in the hindbrain. In
the mouse, loss of function of Hoxa-1 leads to abnormalities
Ectopic expression of Hoxa-1
in the caudal hindbrain. This affects the formation of the
sensory and motor neurons in this region of the neural tube and
the crest normally derived from this region of the hindbrain. In
contrast to an effect on r4-r6, overexpression of Hoxa-1 in the
mouse leads to alteration of the anterior hindbrain (Zhang et
al., 1994). In this case, there is a transformation of r2 into r4
as well as alterations to r3, and effects on expression of other
regulatory genes including Hoxb-1. However, this mouse study
did not observe any effect on the eventual fate of the neurogenic crest-derived tissues nor was later mesenchymal crestderived tissues in the pharyngeal region examined. The results
presented here show that both are dramatically affected by
Hoxa-1 overexpression in the fish.
A number of new observations relevant to the normal function
of Hoxa-1 have come from the current study. The earliest effect
that we analysed showed an alteration to the expression of krx20 in r3 but not r5 in contrast to the situation in the mouse where
no effect is reported on krox-20 in Hoxa-1 overexpressed
embryos (Zhang et al., 1994). In experimental fish embryos, r3
appeared to be enlarged causing the neural keel to become
distorted. This excess growth suggests that Hoxa-1 may be
involved in the control of growth within the neural plate.
Another consistent phenotype is the absence of the first pharyngeal arch skeleton where Meckel’s cartilage and the palatoquadrate fail to form. The second arch-derived ceratohyals are
partially duplicated, most commonly appearing as thickened
elements bifurcated at their medial extent. The transformation
does not affect the more caudally located gill arch skeletal structures but does affect parts of the neurocranium in which the
middle part of the trabeculae are severely malformed indicating
that this region of the trabeculum is derived from the first arch
crest. Furthermore, the most anterior region where the trabeculae fuse to form the ethmoid plate is unaffected suggesting a
separate crest origin for this structure. This is consistent with the
fate map of the neural crest in the chick embryo (Couly et al.,
1993). These results indicate that the neural crest that normally
populates the first pharyngeal arch has to some extent been
respecified. In the normal zebrafish embryo, the first arch neural
crest comes predominantly from r1/2/3, whereas that populating
arch two comes from r3/4/5 (Schilling and Kimmel, 1994).
There are two possible explanations for the abnormality seen
in the experimental embryos; the neural crest from the r2 region
could either (i) enter the first arch but form hyoid rather than
mandible when they differentiate and fuse with the hyoid formed
from the second arch or (ii) migrate incorrectly and end up in
the second arch where they are patterned accordingly and form
additional hyoid tissue. It appears from the analysis of dlx 2
expression (Fig. 10) that the streams of hindbrain crest generating the first and second arches in the wild type are fused into a
single large stream in the Hoxa-1-injected fish. We do not know
what causes this fusion of anterior hindbrain streams but this
maybe due to the abnormal development of r3, which normally
lies between them. It may also be the case that the neural crest
cells within the single large stream may maintain their original
anterior-posterior arrangement and do not intermingle.
Similarly, we still do not know at what stage the neural crest patterning process occurs. Schilling and Kimmel (1994) suggested
that crest cells are determined as to their fate (cell type) prior to
migration; however, they did not comment as to the mechanisms
controlling the eventual patterning process. It is also possible
that RA and Hoxa-1 cause an alteration in the formation of the
743
arches. Such an alteration could reflect ectopic expression of
Hoxa-1 in the hypoblast and could involve the differentiation of
the pharyngeal pouches which develop from endoderm and
separate the arches. This interpretation is consistent with the
observation in rodent embryos that RA leads to the formation of
fused first and second arches (Goulding and Pratt, 1986) into
which neural crest from both r2 and r4 migrate as a single stream
in which r2-derived crest populates the anterior half and the r4derived crest the posterior half (Lee et al., 1995).
Some of our results are similar while others contrast with
the results of a similar Hoxa-1 overexpression study in the
mouse (Zhang et al., 1994). The striking similarity is the transformation or partial transformation of r2 to r4. This was
assessed in our study with respect to the reticulospinal complex
whereas the mouse study analysed this aspect of patterning
relative to a characteristic population of r4 motor neurons. It
would be very interesting in both systems to learn more of the
nature of the transformation in terms of other groups of nuclei.
There were specific differences however, including the absence
in the transgenic mice of any effect on patterning of the Vth
and VIIth ganglion, lack of fusion of the first two pharyngeal
arch derivatives and Krox 20 expression was normal. The
primary data for respecification of anterior to posterior neural
crest in the mouse depends upon alterations to expression
patterns of patterning genes such as Hoxa-2 and Hoxb-1. This
study thus presents no morphological data suggesting a respecification other than the r4 motor neuron population in the
neural ectoderm. It is unclear why these differences exist in the
transgenics because it is likely that all cells express Hoxa-1;
however, there is no indication of the amount of aberrant
protein and timing of expression. In contrast, in the zebrafish,
due to the mosaic distribution of the injected RNA, embryos
will vary in their phenotype. However, it is likely that the RNA
is expressed very early and is still broadly expressed during the
tail bud stages and it is possible that Hoxa-1 protein is
expressed at high levels in the cells containing the RNA.
The results presented in this paper also shed further light on
the association between RA and Hoxa-1. It is evident that the
phenotypes generated by either treating the early zebrafish
gastrula with a pulse of RA (Hill et al., 1995) or ectopic
expression of Hoxa-1 give a phenotype that is strikingly similar
suggesting that there maybe a mechanistic link between RA
and Hoxa-1 function during hindbrain patterning. It is also
established that treatment with RA causes ectopic expression
of Hoxa-1. However, there are a number of unanswered
questions about this possible functional link. Thus, irrespective
of whether Hoxa-1 expression is altered by RA treatment or by
RNA injection, it is not clear when during development ectopically expressed Hoxa-1 has its effect. It could be during early
gastrula stages when transcripts are evident in the shield
following RA treatment (see Fig. 4A-D) or it could be at later
gastrula or early neurula stages when ectopic Hoxa-1 is exclusively in the hypoblast (Fig. 4E-F). If it is at the former, earlier,
stage it could be affecting the epiblast directly. The region of
the epiblast in the midline where ectopic Hoxa-1 is evident
after RA treatment is, at the early gastrula stages, fated to be
midbrain or diencephalon (Woo and Fraser, 1995). Consistent
with the possibility that Hoxa-1 may affect epiblast directly in
the gastrula is that the normal, more lateral, expression of
Hoxa-1 (Fig. 3A-B) is in a region of the epiblast fated to be
hindbrain at 6 hours of development (Woo and Fraser, 1995).
744
D. Alexandre and others
Fig. 7. Duplication of the Mauthner neuron occurs in r2 in Hoxa-1injected embryos. (A-C) 24 hour old embryos stained with the 3A10
antibody. In control embryos (A), a single pair of Mauthner cells is
evident in r4 (arrowheads). In injected embryos (B,C), an additional
Mauthner cell is seen more rostrally. Anterior is up in A and B and to
the left in C. (D,E) Confocal images of hindbrains from larvae that
have had lysinated rhodamine dextran crushed onto their spinal cords
in order to retrogradely label reticulospinal neurons. (D) Control
with the different rhombomeres (numbered) evident from the
segmental arrangement of the neurons. The Mauthner neurons is
arrowed. Duplicated Mauthner cells are evident in r2 in the injected
embryo shown in E. o, otic vesicle; e, eye. Scale bar, 100 µm.
If the effect of Hoxa-1 ectopic expression does not occur until
later in gastrulation, it must be as a result of influencing the
neural plate indirectly through the hypoblast. This must, presumably, be achieved by Hoxa-1 having a down effect on the
transcription of a member or members of an intercellular signalling cascade. A further unanswered question relates to the
functional relationship between RA and other member of the
labial class of Hox genes. The Hoxa-1 paralogues, b-1 and d1 may be ectopically expressed in response to RA treatment in
the zebrafish. It is known that the transcription of Hoxb-1 is
affected by RA in the mouse embryo (Conlon and Rossant,
1992; Marshall et al., 1992) and that mouse Hoxb-1 has two
RAREs in its regulatory sequences (Marshall et al., 1994;
Studer et al., 1994). The zebrafish Hoxb-1 cDNA has yet to be
isolated but analysis of ectopic expression of the mouse protein
in the zebrafish is possible and would be a most interesting
approach with respect to the role of paralogues.
As is the case with RA treatment, injection of Hoxa-1 RNA
results in the respecification of Mauthner neurons leading to
the formation of a hybrid r2 in that not all of the neurons of
the reticulospinal complex are altered. The formation of a
hybrid rhombomere suggests that Hoxa-1 may be affecting
Fig. 8. Hoxa-1 overexpression causes abnormal development of the
head skeleton as shown in ventral views of alcian blue-stained
larvae. (A) Control larva. The Meckel’s (mc) and palatoquadrate (pq)
cartiliages are formed from the first pharyngeal arch mesenchyme.
The ceratohyals (ch) and hyosymplectics (hs) are formed from the
second arch mesenchyme. The gill arches (g) are shown by numbers.
(B) A larva which was injected with Hoxa-1 RNA showing abnormal
development of the jaws. The first arch derivatives are absent and a
broadened pair of ceratohyals are formed (large arrowheads). Small
arrowheads indicate the tip of the abnormally formed trabeculae.
(C) Control larva showing the ventral neurocranium. The anterior
ethmoid plate (ep), trabeculae (tr) and posterior parachordals (pc) are
evident. (D) Dorsal view of the same embryo as in (B). The
trabeculae are malformed (white arrowheads), the ethmoid cartilages
exist as separate pieces anteriorly (black arrowheads) and the
parachordals are normal. (E) A larva that was treated with RA at the
onset of gastrulation showing the abnormally formed jaws. The
forked ceratohyals are arrowed. (F) Two larvae showing the
formation of the jaws at 4 days. The top one has been injected with
Hoxa-1 and lacks the protruding jaws evident in the control larva
below. Scale bar, 100 µm (A-E) or 1 mm (F).
only neurons born early during gastrulation, such as the
Mauthner neuron (Mendelson, 1986). The alternative explanation is that fish Hoxa-1 is not normally expressed in all cells
of r4, a possibility that is supported by the in situ analysis (see
Fig. 3M). In addition, RA treatment and Hoxa-1 overexpression lead to duplication of the Mauthner cell in r4 (see Fig. 2A
in Hill et al., 1995). It is of interest to establish which r2 cells
are transformed into the Mauthner neuron in Hoxa-1-injected
embryos because it has been suggested by Metcalfe et al.
(1986) that reticulospinal neurons exist in specific classes in
Ectopic expression of Hoxa-1
745
each rhombomere. It is possible therefore that cells of each
class are specified by particular genes, such as Hox genes and
overexpression of such genes in an abnormal rhombomere will
only affect cells of that class. The association between RA and
Hoxa-1 makes it a possibility that localised concentrations of
RA act as the trigger for the initiation of Hoxa-1 expression in
the normal embryo. To prove this one needs to know where
RA or other retinoids are in the developing zebrafish embryo
prior to gastrulation at the stage that the endogenous Hoxa-1
gene is first expressed.
Fig. 9. The preotic cranial ganglia are abnormally formed in Hoxa-1injected embryos. This is seen by comparing 24 hour embryos
stained with anti-acetylated tubulin to reveal the VIIth and Vth
ganglia between the otic vesicle (o) and the eye (e). In control
embryos (A), these ganglia are distinct but, in embryos injected with
Hoxa-1 (B), they are reduced and fused into one ganglion
immediately anterior to the otic vesicle. Scale bar, 100 µm.
Fig. 10. (A) A 16 somite control embryo probed for expression of
dlx2. Anterior to the left and dorsal up. Three streams of neural crest
cells can be seen on the lateral regions of the neural keel two of which
lie anterior to the otic placode (op). The anteriormost stream migrates
into the mandibular arch and the preotic stream into the hyoid arch
(see Akimenko et al., 1994). (B) dlx2 expression in a 16-somite-staged
embryo injected with Hoxa-1. Only one preotic stream of crest is now
visible and it is larger than in controls. This aberrant stream of crest
lies immediately anterior to the otic placode and the most anterior
stream of crest is absent. (C) The positioning of the aberrant stream is
confirmed in embryos probed for both dlx2 (blue) and krx-20. (red). In
this 16-somite example, the single preotic stream stretches round the
margin of the abnormally formed r3. Scale bar, 100 µm.
It is a pleasure to thank Robb Krumlauf for his help and encouragement and for providing the mouse Hoxa-1 clone, which was originally isolated by Baron et al. (1987); to Steve Wilson for his interest
and comments on the manuscript and to Monte Westerfield for the dlx
2 probe. This work was supported by grants from BBSRC, MRC and
the Human Frontiers Research Program to N. H., who is a BBSRC
senior research fellow and by a grants from the Wellcome Trust to T.
J. and J. C.; Y. Y. was supported by grants from NIH (1P01HD22486
and 1R01 RR10715) to Professor John Postlethwait. D. A. was
supported by an EMBO long-term fellowship.
REFERENCES
Akimenko, M-A., Ekker, M., Wegmer, J., Lin, W. and Westerfield, M.
(1994) Combinatorial expression of three zebrafish genes related to Distalless: Part of a homeobox gene code for the head. J. Neurosci. 14, 3475-3486
Baron, A., Featherstone, M., Hill. B., Hall, A., Galliot, B. and Duboule, D.
(1987) Hox 1-6, a new mouse homeobox containing gene member of the
Hox-1 complex. EMBO J. 6, 2977-2086
Carpenter, E., Goddard, J., Chisaka, O., Manley, N. and Cappechi, M.
(1993) Loss of Hoxa-1 (Hox1.6) function results in reorganisation of the
murine hindbrain. Development 118, 1063-1075
Chen, Y. and Solursh, M. (1994) A concentration gradient of retinoids in the
early Xenopus laevis embryo. Dev. Biol. 161, 70-76
Chisaka, O., Musci, T. and Capecchi, M. (1992) Developmental defects of
the ear, cranial nerves and hindbrain resulting from targetted disruption of the
mouse homeobox gene Hox 1.6. Nature 355, 516-520
Conlon, R.A. and Rossant, J. (1992). Exogenous retinoic acid rapidly induces
anterior ectopic expression of murine Hox-2 genes in vivo. Development
116, 357-368.
Couly, G., Coltey, P. and Le Douarin, N. (1993) The triple origin of the
vertebrate skull in higher vertebrates - A study in Quail-Chick chimaeras.
Development 117, 409-429.
Duboule, D., Galliot, B., Baron, A. and Featherstone, M. S. (1988). Murine
homeo-genes: some aspects of their organisation and structure. In Cell to
Cell Signals in Mammalian Development (ed. S. deLaat, J. G. Bluemink
and C. L. Mummery), pp. 97-108. Berlin: Springer Verlag, NATO ASI
Series.
Durston, A.J., Timmermans, J., Hage, W.J., Hendriks, H.F., de Vries, N.,
Heideveld, M. and Nieuwkoop, P. (1989). Retinoic acid causes an
anteroposterior transformation in the developing central nervous system
Nature 340, 140-144.
Frohman, M. A. and Martin, G. R. (1989). Rapid amplification of cDNA ends
using nested primers. Technique 1, 165-170.
Frohman, M. A., Boyle, M. and Martin, G. R. (1990). Isolation of the mouse
Hox-2.9 gene; analysis of embryonic expression suggests that positional
information along the anterior-posterior axis is specified by mesoderm.
Development 110, 589-607.
Frohman, M.A. and Martin, G. (1992a). Isolation and spatial analysis of
Hoxd-1, a new murine labial-like gene, reveals that labial subfamily
members are expressed similarly onbly in early anteroposterior axis
formation. Mech. Develop. 38, 55-58.
Frohman, M. A. and Martin, G. R. (1992b). Isolation and analysis of
embryonic expression of Hox-4.9, a member of the murine labial-like gene
family. Mech. Develop. 38, 55-67.
Gendron-Maquire, M., Mallo, M., Zhang, M. and Gridley, T. (1993) Hoxa2 mutant mice exhibit homeotic transformation of skeletal elements derived
from the cranial neural crest. Cell. 75, 1317-1331
746
D. Alexandre and others
Goulding, E. and Pratt R. (1986) Isotreteoin teratogenicity in mouse whole
embryo culture. J. Craniofac. Genet. Dev. Biol. 6, 99-112
Griffin, K., Patient, R. and Holder, N. (1995) Analysis of FGF function in
normal and no tail zebrafish embryos reveals separate mechanisms for the
formation of the trunk and the tail. Development 121, 2983-2994
Hatta, K. (1992) Role of the floor plate in axonal patterning in the zebrafish
CNS. Neuron 9, 629-642.
Hauptmann, G. and Gerster, T. (1994) Two colour whole-mount in situ
hybridisation to vertebrate and Drosophila embryos. Trends Genet., 10, 266.
Hill, J., Clarke, J.D.W., Vargesson, N., Jowett, T. and Holder, N. (1995)
Exogenous retinoic acid causes alterations in the development of the
hindbrain and midbrain of the zebrafish embryo including positional
respecification of the Mauthner neuron. Mech. Develop. 50, 3-16.
Hogan, B., Thaller, C. and Eichele, G. (1992) Evidence that Henson’s node is
a site of retinoic acid synthesis. Nature 359, 237-241
Holder, N. and Hill, J. (1991). Retinoic acid modifies development of the
midbrain-hindbrain border and affects cranial ganglion formation in
zebrafish embryos. Development 113, 1159-1170.
Jegalian, B. and DeRobertis, E. (1992) Homeotic transformations in the
mouse induced by overexpression of a human Hox3.3 transgene. Cell 71,
901-910
Jowett, T. and Lettice, L. (1994) Whole mount in situ hybridizations on
zebrafish embryos using a mixture of digoxigenin and flourescein labelled
probes. Trends Genet. 10, 73.
Kessel, M. (1993) Reversal of axonal pathways from rhombomere 3 correlates
with extra Hox expression domains. Neuron 10, 379-393.
Kessel, M., Balling, R. and Gruss, P. (1990) Variations of cervical vertebrae
after expression of a Hox-1.1 transgene in mice. Cell 61, 301-308
Kessel, M. and Gruss, P. (1991) Homeotic transformations of murine
vertebrae and concomitant alteration of Hox codes induced by retinoic acid.
Cell 67, 89-104.
Kimmel, C.B., Powell, S.L. and Metcalfe, W. (1982). Brain neurons which
project to the spinal cord in young larvae of the zebrafish J. Comp. Neurol.
205, 112-127.
Kimmel, C.B., Metcalfe, W. and Schabtach, E. (1985). T reticular
interneurons: A class of serially repeating cells in the zebrafish hindbrain. J.
Comp. Neurol. 233, 365-376.
Krauss, S., Johansen, T., Korzh, V. and Fjose, A. (1991) Expression of the
zebrafish paired box gene pax (zf-b) during early neurogenesis. Development
113, 1193-1206
Krieg, P. and Melton, D. (1984) Funtional messenger RNAs are produced by
SP6 in vitro transcription of cloned cDNAs. Nucl. Acids Res. 12, 7057-7070.
Krumlauf, R. (1994) Hox genes in vertebrate development. Cell 78, 191-201
Langille, R. and Hall, B. (1987) Development of the head skeleton of the
Japanese Medaka. J. Morph. 193, 135-158.
Langille, R. and Hall, B. (1988) The role of the neural crest in the development
of the cartilagenous cranial and visceral skeleton of the Medaka. Anat.
Embryo. 177, 297-305
Langston, A.W. and Gudas, L. (1992) Identification of a retinoic acid
responsive enhancer 3’ of the murine homeobox gene Hox1.6. Mech. Dev.
38, 217-228
LaRosa, G. J. and Gudas, L. (1988). Early retinoic acid-induced F9
teratocarcinoma stem cell gene ERA-1: alternate splicing creates transcripts
for a homeobox-containing protein and one lacking the homeobox. Mol. Cell.
Biol., 8, 3906-3917.
Lee, Y., Osumi-Yamashita, N., Ninomiya, Y., Moon, C.K., Eriksson, U. and
Eto, K. (1995) Retinoic acid stage-dependently alters the migration pattern
and identity of hindbrain neural crest cells. Development 121, 825-837
LeDouarin, N., Dupin, E. and Ziller, C. (1994) Genetic and epigenetic control
in neural crest development. Curr. Biol. 4, 685-695
Lufkin, T., Dierich, A., LeMur, M., Mark, M. and Chambon, P. (1991)
Disruption of the Hox1.6 homeobox gene results in defects in a region
corresponding to its rostral domain of expression. Cell 66, 1105-1119
Lufkin, T., Mark, M., Hart, C., Dolle, P., Lemeur, M. and Chambon, P.
(1992) Homeotic transformation of the occipital bones of the skull by ectopic
expression of a homeobox gene. Nature 359, 835-840
Mark, M., Lufkin, T., Vonesch, J-L., Ruberte, E., Olivio, J-O, Dolle, P.,
Gorry, P., Lumsden, A. and Chambon. P. (1993) Two rhombomeres are
altered in Hoxa-1 mutant mice. Development 119, 319-338.
Marshall,H., Nonchev, S., Sham, M.H., Muchamore, I., Lumsden, A. and
Krumlauf, R. (1992). Retinoic acid alters hindbrain Hox code and induces
transformation of rhombomeres 2/3 to 4/5 identity. Nature 360, 737-741.
Marshall, H., Studer, M., Popperl, H., Aparicio, S., Kuroiwa, A., Brenner,
S. and Krunlauf, R. (1994) A conserved retinoic acid response element
required for early expression of the homeobox gene Hoxb-1. Nature 370,
567-571
McGinnis, W. and Krumlauf, R. (1992) Homeobox genes and axial
patterning. Cell 68, 283-302
Mendelson, B. (1986). Development of reticulospinal neurons of the zebrafish.
I. Time of origin. J. Comp. Neurol. 251, 160-171.
Metcalfe, W., Mendelson, B. and Kimmel, C. B. (1986) Segmental
homologues among reticulospinal neurons in the hindbrain of the zebrafish
larva. J. Comp. Neurol. 251, 147-159
Morriss, G. (1972) Morphogenesis of the malformations induced in rat
embryos by maternal hypervitiminosis A. J. Anat. 113, 241-250
Morriss-Kay, G. M., Murphy, P., Hill, R. E. and Davidson, D. (1991).
Effects of retinoic acid excess on expression of Hox-2.9 and Krox-20 and on
morphological segmentation in the hindbrain of mouse embryos. EMBO J.
10, 2985-2995.
Murphy, P. and Hill, R. (1991) Expression of the mouse labial-like
homeobox-containing genes, Hox 2.9 and Hox 1.6, during segmentation on
the hindbrain. Development 111, 61-74
Oxtoby, E. and Jowett, T. (1993) Cloning of the zebrafish krox-20 gene (krx20) and its expression during hindbrain development. Nucl. Acid Res. 21,
1087-1095
Papalopulu, N., Clarke, J. D. W., Bradley, L., Wilkinson, D., Krumlauf, R.
and Holder, N. (1991). Retinoic acid causes abnormal development and
segmental patterning of the anterior hindbrain in Xenopus Embryos.
Development 113, 1145-1158.
Postlethwait, J., Johnson, S., Midson, C., Talbot, W., Gates, M., Ballinger,
E., Africa, D., Andrews, R., Carl, T., Eisen, J., Horne, S., Kimmel, C.,
Hutchinson, M., Johnson, M, and Rodriguez, A. (1994) A genetic linkage
map for the zebrafish. Science 264, 699-702.
Ramirez-Solis, R., Zheng, H., Whiting, J., Krumlauf, R. and Bradley, A.
(1993) Hoxb-4 (Hox-2.6) mutant mice show homeotic transformation of a
cervical vertebra and defects in the closure of the sternal rediments. Cell 73,
279-294.
Rijli, F., Mark, M., Lakkaraju, S., Dierich, A., Dolle, P. and Chambon, P.
(1993) A homeotic transformation is generated in the rostral branchial region
of the head by disruption of Hoxa-2, which acts as a selector gene. Cell 75,
1333-1349
Schilling, T. and Kimmel, C. (1994) Segment and cell type lineage restrictions
during pharyngeal arch development in the zebrafish embryo. Development
120, 483-494.
Simeone, A., Acampora, D., Arcioni, L., Andrews, P., Boncinelli, E. and
Mavillo, F. (1990) Sequential activation of Hox2 homeobox genes by
retinoic acid in human embryonal carcinoma cells. Nature 346, 763-766
Simeone, A., Avantaggiato, V., Moroni, M., Mavilio, F., Arra, C., Cotelli,
F., Nigro, V. and Acamporo, D. (1995) Retinoic acid induces stage-specific
antero-posterior transformation of rostral central nervous system. Mech.
Develop. 51, 83-98.
Sive, H. L., Draper, B.W., Harland, R. and Weintraub, H. (1990).
Identification of a retinoic acid sensitive period during primary axis
formation in Xenopus laevis Genes Dev. 4, 932-942.
Sive, H. L. and Cheng, P. F. (1991) Retinoic acid perturbs the expression of
Xhox.lab genes and alters mesodermal determination in Xenopus laevis
Genes Dev. 5, 1321-1332
Studer, M., Popperl, H., Marshall, H., Kuroiwa, A. and Krumlauf, R.
(1994) Role of a conserved retinoic acid response element in rhombomere
restriction of Hoxb-1. Science 265, 1728-1732
Sundin, O. H., Busse, H. G., Rogers, M. B., Gudas, L. J. and Eichele, G.
(1990). Region specific expression in early chick and mouse embryos of
Ghox-lab and Hox 1.6 vertebrate homeobox containing genes related to
Drosophila labial. Development 108, 47-58.
Sundin, O. H. and Eichele, G. (1990). A homeo domain protein reveals the
metameric nature of the developing chick hindbrain. Genes Dev. 4, 12671276.
Westerfield, M. (1995) The Zebrafish Book. Univ. Oregon Press.
Woo, K. and Fraser, S. E. (1995). Order and coherence in the fate map of the
zebrafish nervous system. Development 121, 2595-2609.
Zhang, M., Kim, H-J., Marshall, H., Gendron-Maguire, M., Lucas, D.,
Baron, A., Gudas, L., Gridley, T., Krumlauf, R. and Grippo, J. (1994)
Ectopic Hoxa-1 induces rhombomere transformation in mouse hindbrain.
Development 120, 2431-2442.
(Accepted 13 December 1995)