Lciiio
D
ELSEVIER
Mechanismsof Development 54 (1996) 9-21
Teleost HoxD and HuxA genes: comparison with tetrapods
and functional evolution of the HOXD complex
Frank van der Hoevena,1,Paolo Sordinoaql,Nadine Fraudeaua,
Juan-Carlos Izpistia-Belmonteb, Denis Duboulea,*
aDepurtment
of Zoology and Animal Biology, University of Geneva, Sciences Ill, Quai Ernest Ansermet 30, I21 I Geneva 4, Switzerland
bThe Salk Institute, San Diego, CA 92186-5800, USA
Received 20 June 1995;revision received 24 August 1995; accepted 28 August 1995
Abstract
In tetrapods, Hux genes are essential for the proper organization and development of axial structures. Experiments involving Hox
gene inactivations have revealed their particularly important functions in the establishment of morphological transitions within
metameric series such as the vertebral column. Teleost fish show a much simpler range of axial (trunk or appendicular) morphologies,
which prompted us to investigate the nature of the Hex system in these lower vertebrates. Here, we show that iish have a family of Hox
genes, very similar in both number and general organization, to that of tetrapods. Expression studies, carried out with HoxD and HOXA
genes, showed that all vertebrates use the same general scheme, involving the colinear activation of gene expression in both space and
time. Comparisons between tetrapods and fish allowed us to propose a model which accounts for the primary function of this gene
family. In this model, a few ancestral Hox genes were involved in the determination of polarity in the digestive tract and were further
recruited in more elaborate axial structures.
Keywords:
Hox
genes; Tetrapods; Fish; HOXD complex
1. Introduction
In higher vertebrates, such as rodents or birds, about
40 genes contain a homeobox sequence related to that of
the Drosophila Antp gene. Such Hox genes encode transcription factors and are clustered in four complexes.
Each complex contains from 9 to 11 genes, regularly
spaced over about 200 kb, and transcribed from the same
DNA strand (reviewed in McGinnis and Krumlauf, 1992).
Sequence analyses have revealed that vertebrate HOX
complexes can be aligned with the Drosophila Antennapedia and Bithorax complexes (ANT-C; BX-C) of homeotic genes, indicating an ancient and common phylogenetic origin of this genetic system (Duboule and Doll&
1989; Graham et al., 1989). Thus, during evolution, an
ancestral complex was either split, in Diptera, or amplified along the lineage leading to vertebrates. This large
* Corresponding author. Tel.: +41 22 7026771;
e-mail: duboule@&a.unige.ch.
’ Should be considered equal first authors.
fax: +41 22 7026795;
0925-4773/96/$15.00
0 1996 Elsevier Science Ireland Ltd. All rights reserved
SSDI 0925-4773(95)00455-A
scale amplification may have occurred in parallel with the
emergence of novel vertebrate specialities since cephalochordates appear to have a single complex of HoxZHOM
genes (Garcia-Femandez
and Holland, 1994).
The relationship between this particular genomic organization and the expression patterns of these genes
during fetal development
has been extensively studied.
Briefly, genes located at the 3’ extremities of the complexes (starting with group 1) are expressed from very
anterior positions, i.e. within the developing hindbrain,
while genes located at 5’ positions (e.g. group 12 or 13
genes) are expressed in progressively
more restricted
posterior areas. This ‘spatial colinearity’, which was first
observed genetically in Drosophifa by Lewis (1978), and
subsequently documented at the molecular level, is also
observed in vertebrates (Gaunt et al., 1988) together with
another type of colinearity,
or ‘temporal colinearity’,
which describes the temporal progression
with which
these genes are activated during development,
such that
3’-located genes are expressed before by their S-located
neighbours in the complex (Doll6 et al., 1989, Izpisda-
10
F. van der Hoeven et al. I Mechanismsof DevelopmentS4 (I 996) 9-21
Belmonte et al., 1991). This observation suggested that
the temporal sequence of Hex gene activation is causally
linked to the physical ordering of the genes along their
complexes (Duboule, 1992).
The importance of vertebrate Hox genes in the organization of the body plan is now well documented. Experiments involving either gain or loss of function have revealed that they are required to properly build (identify)
structures (e.g. Krumlauf, 1994). The absence of a given
HOX product will generally result in the transformation of
a structure into a similar, but different, structure from the
same anatomical series (e.g. a lumbar into thoracic-like
vertebra), transformations
which are often referred to as
homeotic (e.g. LeMouellic et al., 1992; Doll6 et al., 1993;
Ramirez-Solis
et al., 1993; Rijli et al., 1993; Kostic and
Capecchi, 1994). From these results, it appears that different anterior-posterior
levels of the body are defined by
the presence or absence of given Hox proteins. In addition, the levels of the respective proteins found at a given
position are important for the proper development of the
concurrent structure (e.g. Condie and Capecchi, 1994).
Hence, changes in the distribution of these proteins during development might lead to variations in the respective
importance of the different body parts with respect to
each other, a phenomenon
referred to as ‘transposition’
(Goodrich,
1913). Consequently,
modifications
of the
mechanism(s) behind colinearities might have been a rich
source of morphological
variations
during evolution
(Duboule, 1994; Burke et al., 1995).
Tetrapods can exhibit very different vertebral formulae
(i.e. number of cervical, thoracic, lumbar, sacral and cauda1 vertebrae) and the transition from one type of vertebral morphology to another type likely involved critical
differences in the distribution of Hox proteins. The relationships between such transitions and the Hox proteins
involved may have been conserved throughout the evolution of vertebrates since mammals and birds appear to
have comparable Hox proteins acting at homologous vertebral transitions (Gaunt, 1994; Burke et al., 1995). However, lower vertebrates, such as agnathans or fishes, do
not show clear transitions in their vertebral morphologies.
Their simpler metameric organization
may thus not require as many Hox proteins as a complex tetrapod vertebral column does.
To investigate the evolutionary
relationships between
Hox proteins and variations of the vertebrate body plans,
we characterized the posterior Hox genes in the teleost
zebrafish (Danio rerio). Here, we report on the cloning,
organization and expression studies of gene members of
either the HoxD or HoxA complexes, and propose the
following conclusions:
(1) fish have at least four Hox
complexes, the structures of which seem to be very similar to those of tetrapods; (2) the functional organization of
the fish HoxD complex (i.e. both spatial and temporal
colinearities)
is comparable
to that observed during
tetrapod development;
(3) a transient combinatorial
dis-
tribution of Hox proteins occurs in the fish trunk mesoderm even though no important morphological
vertebral
variation is observed; (4) the extent of the ‘Hox morphological window’ seems to be fixed by the position, in
space and time, of the analia, since as in tetrapods, the last
Hox gene to be activated is strongly expressed in the presumptive analia-genitalia
area. From this, we conclude
that the function of Hox genes in the establishment of the
vertebral formulae, in higher vertebrates, is a late acquisition which was made possible by the pre-existence
of
different exprescion domains along the AR axis. We
speculate that the original deployment of ancestral Hox
genes from anterior to posterior was linked to the necessity to position and differentiate the most posterior part of
the digestive tract. This genetic system may have been
recruited later in more complex animals, in the organization of both the axial nervous system and somitic mesoderm.
2. Results
2.1. The zebrafish HoxD complex
Screening of both cDNA and genomic libraries led us
to isolate a piece of Danio DNA of about 40 kb containing the upstream (‘posterior’) part of the fish HoxD complex. A region starting immediately 5’ of the Hoxd-9 gene
and extending upstream the Hoxd-I3 gene was entirely
sub-cloned, mapped and characterized. Four transcription
units were sequenced, corresponding
to the Hoxd-IO,
Hoxd-II, Hoxd-12 and Hoxd-I3 genes. The general organization of the fish HoxD complex was found to be
very similar to that of higher vertebrates. The overall size
of this part of the complex in fish is similar to that of the
mouse (Fig. 1A). In addition, the relative distances between neighbour genes were conserved. Interestingly, the
intergenic distances were slightly shorter in fish, but this
was compensated by longer intronic sequences (Fig. 1A).
As in tetrapods, all genes were transcribed from the same
DNA strand.
Sequencing of various genomic and cDNA clones confirmed that these four genes were orthologous
to the
tetrapod Hoxd genes. In all cases, the complete coding
sequences were determined
and alignments
with the
mouse counterparts revealed different extents of similarities (Fig. 1B). While the fish Hoxd-IO and Hoxd-I1 proteins were highly similar to their murine cognates, the
Hoxd-I2 and, particularly,
the Hoxd-13 proteins were
unexpectedly divergent between lower and higher vertebrates. This observation was particularly obvious when
the N-terminal parts of the proteins were compared, but
this unusual sequence divergence
was already visible
when the homeodomain
sequences alone were considered. While only two amino acid exchanges were observed between either the fish and mouse Hoxd-IO or
Hoxd-11 homeodomains,
a similarity related to that usu-
F. van der Hoeven et al. I Mechanisms of Development 54 (1996) 9-21
A
11
Mouse
HOXD
Fish
s
E
I
Xh Xh Xh
X
E
E
X
1 tb
B
I
mv+10
fix&l
I
Hndl2
Hxd-13
D
MSFPNSSPPANTFLVDSLISACRSDSFYSSSASMYMPPPSADMGTYGMIITCGLLPSW(IPQVDSWTDPNRSCRIEQPVTaQ
*
)o** II*t*..********* . tt*.**
*
*
*
~FPNSSPRANTFLVD-LIGRCRTD-FYSSS-NMYMPARTAEMGTYGMQTCGLLPAffiKRGEVNHHSYIPQTDTWADPSRSCRLEQPL-Nn
l
10
l
l
l
l
l
F
F
F
12
FHad-
12
M
F
M
F
Fig. 1. Posterior part of the Danio HoxD complex and characterization of the Hoxd-IO to Hoxd-13 genes. (A) Schematic comparison of the mouse (top)
and fish (middle) HoxD complexes at the same scale. The relative distances between the genes are conserved. Black boxes indicate exons. The two
bottom lines show maps of the subclones of the four fish Hex genes characterized in this work. Stippled boxes represent the exons and the open rectangles delineate the positions of the RNA probes used for the in situ hybridizations. H, HindIII; E, EcoRI; S, SalI; Xh, XhoI; X, XmaI; Sm, SmuI. (B)
Homeodomain sequences of the fish (bottom lines) and mouse (upper lines) Hoxd-IO to Hoxd-13 genes. The stars indicate amino acid exchanges
among orthologous sequences. (C) Extent of protein sequence similarities between mouse and chick (light grey rectangles) or mouse and fish (dark
grey) Hoxd homeodomains. While all Hoxd homeodomains are essentially identical among higher vertebrates, a significant divergence is observed
between mouse and fish sequences when more posterior genes are considered (see also (B) and (D)). (D) Protein sequence comparison between mouse
and fish Hoxd-IO (upper part) and Hoxd-12 (bottom part). The mouse (M) sequence is on the upper lines. The shaded rectangles contain the homeodomain sequences and the black arrowheads indicate the position of the introns. The stars show identical amino-acids. While the two orthologous
Hoxd-10 proteins are reasonably similar, the Hoxd-I2 similarities are barely detectable in the parts of the proteins encoded by the first exons. (E) Homeodomain sequences of the fish (top) and mouse (bottom) Hoxa-11 gene.
ally obtained within tetrapod species, 10 residues were
different between the fish and mouse Hoxd-I2 homeodomains, a divergence much above the expected number
(Fig. 1B). The Hoxd-13 homeodomains
were even more
divergent as 13 amino acids differed between the fish and
the mouse proteins (Fig. 1B). However, extended se-
F. van der Hoeven et al. I Mechanisms of Development 54 (1996) 9-21
12
quencing within the intronic and intergenic regions demonstrated behind any doubt that this was the genuine fish
HoxD complex.
These unexpected differences between mouse and fish
homologous homeodomain sequences were increasingly
important when genes from the 5’ end of the complex
were considered (Fig. 1C). In contrast, birds and rodents
showed high conservation of their cognate homeodomain
sequences, irrespective of the position of the gene within
the complex (Fig. 1C; see Izpisua-Belmonte et al., 1991).
The divergence between mouse and fish proteins was
even stronger when full protein sequences were considered, as exemplified by the sequence comparisons between the mouse and fish Hoxd-IO and Hoxd-I2 full proteins. While the mouse and fish Hoxd-IO proteins were
well conserved, similar to the conservation observed
among higher vertebrates, the fish Hoxd-I2 protein was
highly divergent from its rodent counterpart (Fig. 1D) and
the d-13 sequences were almost unrelated (not shown).
The potential meaning of this observation in the context
of the evolution of the concurrent functions, in lower and
higher vertebrates is discussed below.
2.2. At leastfour HOX complexes in teleostfish
An additional cDNA clone encoding a paralogy group
11 homeobox was sequenced. The homeodomain sequence suggested that this cDNA clone derived from the
Hoxa-II gene, as only one amino acid exchange was obAbdB abdA Ubx
Bx-c
a-11
a-10
2.3. Spatial and temporal colinearities in thefish trunk
The earliest detection of Hoxd-IO transcripts was recorded along the main body axis, soon after the onset of
somitogenesis. At 11 hpf (4 somites) Hoxd-ZO was expressed in the dorsal pre-somitic mesoderm and in the
neural keel at posterior levels (not shown). Transcripts
encoded by the Hoxd-ZZ, Hoxd-I2 and Hoxd-13 genes,
which appeared sequentially in the trunk at about 12, 14
and 17 hpf, respectively, were similarly located (Fig.
3A,B). Subsequently, while the tail was budding and
elongating, the expression patterns extended into ventral
somitic mesoderm (Figs. 3C,D and 4A,B). At 22 hpf, a
Antp
Scr
Dfd
lab
pb
m---m
II-cFII_
. ..
a-9
served when compared to the mammalian Hoxa-ZZ homeodomain (Fig. IE). DNA sequences from either side of
the homeobox, as well as cloning of the genomic region
confirmed that this clone corresponded to the fish HoxaZZ gene. As for Hoxd-ZZ, the sequence similarity with the
cognate tetrapod homeobox was high. Furthermore, our
screen allowed us to isolate additional homeoboxcontaining cDNAs belonging to different paralogy groups
on different Hox complexes (not shown), and genes
members of either the HoxB (Njolstad et al., 1990) or
HoxC (Ericson et al., 1992) complexes have been previously reported. Altogether, these data demonstrate the
presence of at least four complexes of Hox genes in
teleost fish, a situation similar to that of higher vertebrates
(Fig. 2).
a-7
a-6
a-5
ANT-C
a-3
a-4
a-2
a-l
HOXB
b-9
b-8
c-13
c-12
c-11
c-10
c-9
c-8
h'vx-2 d-13
d-12
d-11
d-10
d-9
d-8
b-7
b-6
b-5
b-4
c-6
c-5
c-4
d-4
b-3
d-3
b-2
b-l
d-l
++
Late/Posterior
4
Early/Anterior
Fig. 2. The Hox gene family in Danio rerio. The scheme represents the structure of the Hox gene family in higher vertebrates. The black boxes indicate
which genes have been isolated to this date in the zebrafish. Members of the four clusters have been reported which indicates that fish have (at least)
the full complement of Hox genes. Data am from this work, Njolstad et al. (1990); Ericson et al., (1993); T. Kondo and PS, unpublished (Hoxa-9,
Hoxa-IO and Hoxa-13) and N. Holder, unpublished (Hoxa-I).
F. van der Hoeven et al. I Mechanisms of Development 54 (1996) 9-21
hierarchy of transcript abundance was visible along the
AP axis, with the strongest accumulation at the posterior
end, in the region of somite proliferation (Fig. 3). Between 22 hpf and 26 hpf, following a rostra1 to caudal
progression, transcript levels decreased in lateral and
ventral mesoderm while expression was maintained in
lateral clusters of cells in the neural rod (Fig. 4B,C). The
rostra1 boundaries of expression along the AP axis were
colinear with the positions of the genes in the cluster (Fig.
5) and the anterior limits of Hoxd-IO, Hoxd-Zl, and
Hoxd-12 transcript domains were assigned to somite levels 9, 11, and 17, respectively, while Hoxd-13, at corresponding stages, was expressed at the posterior tip of the
animal (Fig. 5). By the end of somitogenesis (24 hpf to 26
hpf), the intensity of the Hoxd-20, Hoxd-11 and Hoxd-22
hybridization signals in the trunk began to weaken, while
the more posterior segments became negative (e.g. compare Hoxd-II in Figs. 3C and 5B). In contrast, Hoxd-13
transcripts were progressively restricted to the posterior
end of the somitic mesoderm (Fig. SD). Between 40 and
56 hpf, all four Hoxd genes were strongly downregulated
and their expression in the trunk was undetectable.
Results obtained with the Hoxa-11 gene confirmed that
the expression of fish Hox genes in the developing trunk
followed the same general rules that applied to their
mammalian counterparts. As for Hoxd genes, Hoxa-ll
was activated in the early tail bud and subsequently displayed a clear rostro-caudal boundary. This boundary,
however, was observed at the level of somite 10 (Fig.
3E), i.e. at about the same limit as that of the paralogous
Hoxd-II gene (compare with Fig. 3B). Expression extended more anteriorly in the neural keel than in the ventro-lateral mesoderm (Fig. 3E), a feature already observed
in rodents.
2.4. Expression in hindgut and urogenital sinus
In a way strikingly similar to their rodent cognates,
some Hoxd genes were expressed in the terminal part of
the developing hindgut as well as in the urogenital region.
In particular, Hoxd-13 transcripts were detected at 15 hpf,
i.e. before any expression in the tail bud, as a discrete
transverse stripe of ventral cells (upon lateral view) in the
presumptive area of the proctodeum (Fig. 6A). From the
dorsal side, this discrete stripe was composed of a few
cells which, at this stage, probably labelled the future
position of the cloaca (Fig. 6B). At 22 hpf, the expression
domains of Hoxd-12 and Hoxd-I3 displayed distinct
boundaries and were comprised of all the incipient urogenital bud and mesenchymal cells surrounding the most
posterior aspect of the pronephric duct (which, at this
stage, marks the site where the anus will eventually form).
The expression of these two genes appeared coordinately
regulated and in agreement with colinearity, for the more
5’-located gene, Hoxd-13, was transcribed more intensely
(Fig. 6D, compare with the expression in the trunk) and
13
with a more distal extension (including the entire future
cloaca) than Hoxd-12, which extended more anteriorly in
the hindgut, but not as distally in the cloaca (Fig. 6C, arrows).
After 22 hpf, the expression domains of Hoxd-I2 and
Hoxd-I3 in the distal genitalia and analia underwent important changes. While the expression of Hoxd-I2 was
rapidly downregulated such that no transcripts were seen
at 24 hpf, Hoxd-13 expression remained clearly detectable
until 64 hpf. Hoxd-I3 transcripts became gradually more
restricted toward distal epithelial or muscle cells surrounding the future cloacal region (Fig. 6E,F), a situation
similar to that observed in rodents (Doll& et al, 1991).
3. Discussion
3. I. Teleost fish have at least four HOX complexes
This work reports the cloning and expression of zebrafish Hox genes belonging to two different complexes;
first, the complete 5’ end of the HoxD complex and, secondly, a group 11 gene member of the HoxA complex.
Previous studies have demonstrated the presence of Hox
genes belonging to both the HoxB and HoxC complexes.
Additional genes members of these latter complexes, such
as Hoxb-9, Hoxb-3 and Hoxc-9, as well as Hoxu-13 have
been isolated by either cDNA screening or PCR amplification (our unpublished work). Altogether, the results
indicate that teleost fish have at least four HOX complexes and therefore have the full complement of Hox
genes found in higher vertebrates such as birds or mammals.
As both modern bony fish (e.g. Danio) and high vertebrates (e.g. Mus) have a minimum of four copies of an
ancestral complex, we conclude that more primitive
actinopterygians as well as sarcopterygian fish (did) have
the same genetic configuration. Therefore it appears that
the global genomic amplification leading to the fourcomplex situation must have occurred before, or concomitantly with, the appearance of gnathostomes. The
fact that cephalochordates, such as Amphioxus, seem to
have a single complement of these genes (Holland et al.,
1994) suggests that this important event took place during
chordate evolution, perhaps in conjunction with the appearance of chordates of intermediate complexity such as
agnathan vertebrates (e.g. the lamprey; Pendleton et al.,
1993). The genomic analysis of their HOX complement
would help clarify this possibility. Alternatively, it is
possible that the two- or three-complex configuration was
evolutionarily unstable and rapidly disappeared after further duplication.
In the course of this work, a Hox gene was isolated
which could not be attributed to any of the reported clusters (not shown). While this could result from a punctual
duplication, it might also illustrate a peculiarity of the fish
4
F. van der Hoeven et al. I Mechanisms of Development 54 (19%) 9-21
F. van akr Hoeven et al, I Mechanisms of Development 54 (1996) 9-21
15
Fig. 4. Histological sections showing the distribution of Hoxd-I2 transcripts at three different developmental stages in the trunk. (A) Mid-sag&al section through the tail bud of specimen at 14 hpf, i.e. similar to the one shown in Fig. 3A. A clear anterior boundary is observed (arrowhead) and expression is detected in all cells but the notochord. (B) At a later stage (22 hpf), a cross section reveals that the Hoxd-12 gene is still expressed in a
variety of cell types dorsally, ventrally and laterally. (C) In an older specimen (34 hpf), however, expression is restricted to clusters of centro-lateral
cells in the neural tube (arrowheads). nt, notochord; tb, tail bud; kv, Kupfers vesicle; me, mesoderm; nr, neural rod; ntu, neural tube.
genome, which seems to systematically
contain more
orthologous genes than its higher vertebrate counterparts.
Several gene families in fish thus contain more genes than
their tetrapod counterparts, for example, genes related to
msh (Holland, 1991) or even-skipped (Joly et al., 1993).
So far, one additional copy is always obtained, suggesting
that teleost fish may have experienced
a sub-genomic
amplification
during their evolution. Alternatively,
this
may indicate that early vertebrates had more than four
complexes. This counter-intuitive
possibility would imply
that one complex(es) was lost along the lineage leading
from ancestral fish to higher vertebrates.
3.2. Luxitas terminalis
Even among tetrapods as different as urodeles and
mammals, protein sequences of orthologous Hex open
reading frames are extremely well conserved, both inside
and outside the homeodomains
(e.g. the newt versus
mouse Hoxd-11 genes; Brown and Brockes, 1991; Izpisua-Belmonte
et al., 1991). Such a high conservation
must reflect the persistence, in these animals, of very
similar functions, not only with respect to the DNA
binding specificity (achieved by the homeodomains)
but
also regarding the regulatory functions carried out by
other protein domains. The complete sequences of the
four zebrafish transcription units (from Hoxd-10 to Ho&13), however, gave us a slightly different figure. While
the more ‘anterior’ genes (i.e. Hoxd-IO and, to some extent, Hoxd-11) were still highly similar to their mammalian or avian counterparts, the Hoxd-I2 and, in particular,
Hoxd-13 genes, were very different. Ten differences in
amino acid residues were observed between the Hoxd-12
homeodomain sequences, a divergence far above the one
to two exchanges normally scored in comparable cases
(see e.g. Hoxd-Zl and Hoxd-IO). The N-terminal half of
the fish Hoxd-22 protein was also much less similar to its
mouse counterpart, even though sequence alignment at
Fig. 3. Early expression patterns of the Hoxd-II (A-D) and Hoxa-1Z (E,F) genes. The earliest detection was at 6 somites stage (12 hpf) (A) and the
transcript domain was well established at the early tail bud stage (16 hpf; B). (C) Expression of Hoxd-II in an elongating tail bud (16 hpf) showing the
rather homogeneous labelling throughout the tail bud except for the notochord and the surrounding ectoderm which are negative. The labelling is
reinforced dorsally. (D) At 22 hpf, Hoxd-II is still expressed in the progressing tail bud with maximum intensity distally. (E,F) Two early stages in the
activation of the Hoxa-11 gene. At both 14 hpf (E) and 16 hpf (F), the domains are similar to those of the Hoxd-11 gene, as shown by the positions of
the white arrowheads in (A), (B), (E) and (F) which indicate the anterior positions of the expression domains, but extend slightly more anteriorly. h,
headfold; A, anterior; P, posterior; tb, tailbud; nf notochord; me, mesoderm; ec, ectodenn; c, cloaca.
16
F. van der Hoeven et al. I Mechanism
the N-termini clearly reflected the orthologous relationship between these two genes. In the case of Hoxd-13, the
homeodomain had 13 amino acid exchanges with respect
to the mouse and the upstream part of the protein (as deduced from a complete cDNA) was barely recognizable.
Only a few clusters of amino acid residues were conserved at the N-terminus. We thus observed a clear gradient in the extent of the fish to mouse protein sequence
similarities, so that an increasing divergence was seen in
genes located at increasingly more ‘posterior’ positions in
the complex.
We interpret this observation as an indication that the
set of functions achieved by the most posterior Hox proteins has not been as tightly conserved between lower and
higher vertebrates, as has the set of functions carried out
by ‘anterior’ Hox proteins. In other words, functions carried out by anterior proteins, such as patterning within the
developing hindbrain, are essential and reflect fundamental traits of vertebrates. Accordingly, no tolerance is observed, at this anterior level, for functional variation during evolution, as this would affect major features of the
phyla (e.g. the identities of cranial nerves, migration of
crest cells, etc.). In contrast, the functions of the most
posterior Hox genes may not all be linked to basal vertebrate features. For example, while the function of Hoxd13 in the determination of the proctodeal area is likely of
primary importance for all vertebrates (see below), the
involvement of the mammalian Hoxd-13 gene in either
patterning the autopods, marking the vertebral sacrolumbar transition or patterning the penian bone (Doll6 et
al., 1993; Sordino et al., 1995) probably has no functional
equivalence in fish. Consequently, a concurrent evolutionary decrease in sequence similarities was allowable.
From this, it appears that Hox-dependent important variations in body plan between higher and lower vertebrates,
should be preferentially observed in most posterior parts
or, similarly, in distal parts of the appendicular skeleton
(in all ‘extremities’). In the latter case, this could account
for the previously described principle of ‘proximal stability versus distal variability’ (Hinchliffe, 1991; see
Duboule, 1994). We propose to refer to this posteriorly
(terminal) increased tolerance for functional variability,
paralleled by a higher divergence in Hox protein sequences, as ‘L.axitus terminalis’ (terminal laxism). This
means that, within the frame of the Hox system, late occurring (posterior or distal) structures (such as autopods)
are more susceptible to radical evolutionary changes than
are structures produced earlier and at more anterior positions.
of Development 54 (1996) 9-21
3.3. Axial specification iw teleost fish
Unlike tetrapods, teleost fish do not have a large variety of highly specialized vertebral types. The major difference observed along the rostro-caudal axis is at the
ventro-lateral sides of the vertebrae; while anterior vertebrae articulate with ribs, posterior (post-anal) ones have
smaller pieces which form the haemal arches, surrounding
the caudal vein and artery. No specific vertebral transition
is observed at the level of the pelvic fins. In addition, as
these latter appendages are not fixed to the axial skeleton
by a genuine belt, they can be found at different axial
levels in various species. Also the myomeric structure
appears rather homogenous from anterior to posterior,
with poorly individualized lateral muscles. This rather
simple metameric organization coincides with a very
transient expression of Hoxd genes in trunk mesodermal
derivatives, which supports the view that teleost fish may
not depend heavily on the deployment of multiple Hox
proteins in order to express their rather homogenous vertebral identities.
The fact that the expression of these genes was maintained in the developing spinal chord may be relevant to
their colinear expression along the rostro-caudal axis. In
the neural rod, Hoxd genes were expressed laterally, in
restricted groups of neurons, at a stage. when expression
in mesoderm derivatives was extinguished. It is thus possible that colinear expression of ‘posterior’ Hox genes in
the developing trunk is required to specify differences in
neuronal types along the rostro-caudal axis. In this view,
differential expression in mesoderm derivatives should be
considered either as a vestige of an ancestral configuration, i.e. of a more ancient, non teleost, fish which might
have shown a greater variety of mesodermal structures
along its AP axis. Alternatively, this could simply be a
consequence of the strategy by which these genes are
turned on, in both mesoderm and ectoderm lineages
within the tail bud, with no particular function in the axial
skeleton.
3.4. Posterior Hox genes and the morphogenesis of the
proctodaeum area
Strong expression of Hoxd-I3 was seen in the developing analia-genitalia in a way similar to what has been observed in higher vertebrates (Doll6 et al., 1991). Additionally, mice carrying two Hoxd-13 mutant alleles have
phenotypic alterations of both their genitalia (DollC et al.,
1993) and analia (unpublished). It thus appears that the
Fig. 5. Colinear expression of posterior Hox genes along the developing fish rostra-caudai axis.
Hoxd-10 to Hoxd-I3 genes in ca. 24 hpf developing fishes. The open arrows, on the dorsal sides.
boundaries while the black arrows, ventrally, point to the cloaca to help comparing the different
within the HoxD complex are diagrammed. (E) Expression of the Horn-1 1 gene at a comparable
that of Hoxd-IO (A) and Hoxd-1 I (B). See the text for somite levels.
(A-D) Comparison of the expression domains of the
show the approximate positions of the AP expression
panels. On the right side, the positions of the genes
stage which shows an AP boundary located between
18
F. van der Hoeven et al. I Mechanisms
most posterior vertebrate AM-B-related genes, i.e. the
group 13 genes, are involved in the development of the
analia-genitalia region. This is confirmed, in both mouse
and fish, by the onset of the expression of the Hoxd-13
gene in the primordia of the proctodaeum (see Tucker and
Slack, 1995; this work, Doll6 et al., 1994 and unpublished). In Drosophila, Abd-B is expressed in hindgut
mesoderm and the iab9 mutation of Abdominal B (Abd-B)
affect both genitalia and analia (Celnicker et al., 1990). In
C. elegans, the AbdB-related gene egt-5 has an expression
pattern that characterizes the anal region, in particular in
the rectal blast cells which line the rectum (Chisholm,
1991; Wang et al., 1993).
Expression in the posterior gut was also observed for
the Hoxd-12 gene, though at a weaker level and not in the
most posterior part. This confirms and extends previous
observations made on the colinear distribution of posterior Hoxd transcripts in the developing mouse genital
apparatus (Doll6 et al., 1991). It also reinforces the observation that a posterior gene (e.g. Hoxd-13) is systematically expressed at higher level than its 3’-located neighbours (e.g. Hoxd-II or Hoxd-22) in this area (quantitative
colinearity; Doll6 et al., 1991). Recent studies have
pointed to the colinear expression of posterior Hox genes
in the avian gut (Yokouchi et al., 1995; Roberts et al.,
1995).
3.5. The posterior end of the Hox system
The fact that Hoxd-13 (the last Hox gene) is activated
at the time and position at which the proctodaeum area is
defined, might provide an explanation for the differences
observed in the AP levels of Hox expression between
lower and higher vertebrates. While the position of the
analia, with respect to the metameric series, is indeed
rather fixed posteriorly in amniotes (at around level 2%
30), teleost fish have a cloaca located much more anteriorly (at around level 16-17 in Dunio). Therefore, the
early expression of Hoxd-13 in the presumptive proctodaeum will result in the activation of this gene at the corresponding axial level (somite 16-17). This is well in
agreement with what is reported in this work. In this
view, the large offset, between mouse and fish, in the
levels of expression boundaries for posterior Hoxd genes
of Development 54 (I 996) %21
(at least ten metamers more anteriorly) would be due to
the requirement for the last gene (group 13) to be expressed at the body level at which the proctodeum will
form. This would fix the posterior extent of the Hox system at the level of the analia. Consequently, everything
located posterior to the proctodaeum area (the ‘tail’ ac!cording to Tucker and Slack, 1995) could not benefit
from the activation of novel Hox gene. This may be illustrated by the fact that vertebrate tails, when present, do
not show important morphological transitions.
An important consequence of the association between
the formation of the proctodaeum and the expression of
the group 13 Hox genes is that of the speed of temporal
colinearity. In rodents, the activation of the entire Hox
network takes about 2 days; the time spent between the
formation of the foregut pocket and the proctodaeum. In
Danio, the complete set of Hox genes are activated in a
few hours. While this reflects the different developmental
strategies, it means that the intrinsic structures of the Hox
complexes does not fix the kinetics of activation. We
show here that very similar complexes (same general organization, same length) can be activated within very
different time scales.
We previously reported that the mouse Hoxa-I gene
(formerly Hox-I.6), the most ‘anterior’ gene of the HoxA
complex, is expressed early in the development of the
foregut pocket. We now show that there is a correspondence between the Hoxd-I3 gene and the morphogenesis
of structures deriving from the proctodaeum. We therefore conclude that Hox genes located at the two extreme
positions, within their complexes, are expressed at both
extremities of the digestive tract. We believe that this
important observation reflects the ancestral function of
this gene family. In this scheme, one of the first anterior
to posterior differences which needed to be specified in a
simple, ancestral metazoan was the fixation of the anterior, middle and posterior parts of a primitive digestive
tube. Once established, such a system would have been
recruited in parallel with the functional diversification of
metamers, by both neural cells and mesoderm derivatives,
but always within a morphological window extending
roughly from the mouth to the anus, which seems indeed
to be a recurrent trait in the Hox systems of all animals
analysed so far.
Fig. 6. Expression of the Hoxd-13 (A,B,D-F) and Hoxd-12 (C) genes in the developing analia and urogenital system. (A,B) Expression of Hoxd-13 at
15 hpf. The first signs of labelling appear as a stripe of cells (lateral view in A; dorsal view in B), which encompass the prospective area of the proctodaeum (open arrow in A; arrowheads in B). At later stages (C-F), this domain will be reinforced and match the cloacal region. (C) At 22 hpf, the
Hoxd-I2 gene is expressed in the posterior gut, all along the duct (black arrowheads) but not up to the most distal portion which is negative (between
the open arrowheads). (D) In contrast, the Hoxd-I3 gene, at the same age, is transcribed only in the most distal part of the hindgut (not seen between
the black arrowheads), the future rectum, as well as in all the cloaca up to its distal end (between the open arrowheads; compare with C). The intensities of expression are also different with respect to the expression in the trunk; Hoxd-It is weaker in the cloacal region than in the tail (C) while Hoxd13 is much stronger in the cloaca (D). (E) At 25 hpf, a strong Hoxd-13 signal is detected in the entire cloacal region, However, this signal starts to
weaken and is even more difficult to detect (though still clear) in a specimen at 64 hpf (F). In this latter case, only a few cells located at the distal tip of
the cloca are positive. tb, tail bud; A, anterior; P, posterior; c, cloaca; hg, hindgut; ud, urinary duct; gd, genital duct.
F. van der Hoeven et al. I Mechanisms of Development 54 (1996) 9-21
19
20
F. van der Hoeven et al. I Mechanisms
4. Material and methods
4.1. Fish stocks
Wild type fish (Danio rerio) were originally purchased
at the Famila Center, Heidelberg. In situ hybridization
studies were performed with embryos coming from a
stock of zebrafish homozygous for the mutations golden
(golb’), and sparse (spab*) and heterozygous for albino
(albb4) (Streisinger et al., 1986), kindly provided by Dr. F.
Rosa. This mutant line develops a very weak pigmentation and hence facilitates detection of the staining. Fish
were naturally random-bred and the embryos raised according to Westerfield (1989). Staging of the embryos is
given in hours (h) after fertilization (hpf) with-an incubation temperature of 285°C.
4.2. isolation of zebrafish Hex complex genes
Initially, a zebrafish cDNA library (Fjose et al., 1994,
gift of Dr. C. Fromental-Ramain) was screened with 32Plabelled mouse genomic DNA fragment containing a variety of Hoxd gene homeobox sequences. Several clones
containing homeobox sequences belonging to different
complexes were isolated, among which was a partial
cDNA clone corresponding to the fish Hoxd-Zl gene.
This clone was further used to screen a genomic library
@fix11 Stratagene). A phage was obtained and used as
starting point for walking in both directions. Several
overlapping phages were isolated and characterized as
covering a piece of genomic DNA starting at approximately 20 kb 5’ from the Hoxd-I3 gene and extending up
to the start of the Hoxd-9 transcription unit. The coding
sequence of the four genes (Hoxd-IO to Hoxd-13) were
established and can be found in the EMBL database under
the accession numbers X87750 (Hoxd-ZO), X87751
(Hoxd-ll), X87752 (Hoxd-23) and X87753 (Hoxd-Z2).
Similarly, a cDNA clone corresponding to the Hoxu-11
gene was isolated and sequenced. Sequences upstream
and downstream the homeobox and comparison with the
mammalian cognate genes confirmed the attribution of
this group 11 gene to the HoxA complex (not shown).
This clone was further used to isolate a corresponding
phage on the HoxA complex.
4.3. RNA probes
The following genomic DNA fragments were used to
synthesize antisense RNA probes for whole-mount in situ
hybridizations: Hoxd-10; a 320 bp XbaI-Hi&III fragment
(XH 320) partially spanning the homeobox. Hoxd-I 1; a
570 bp EcoRI fragment (4.60) partially spanning the homeobox. Hoxd-Z2; a 1100 bp EcoRI-XhoI fragment
(EX900) spanning the homeobox. Hoxd-13; E1900, a
275 bp EcoRI fragment partially spanning the homeobox.
The clones were linearized with XbaI, HindIII, EcoRI and
of Development S4 (1996) 9-21
SacI, respectively. For Hoxu-II, a 1.3 kb EcoRI cDNA
fragment containing the entire homeobox plus 3’-located
sequences was used. Anti-sense RNAs were generated
with the T7 and T3 RNA polymerases in the presence of
digoxigenin-labelled UTP according to Boehringer’s
specifications. In addition, the probes were passed
through a NucTrap@ push column (Stratagene) before
ethanol precipitation and fractions were analysed on an
agarose gel.
4.4. Whole Mount in situ hybridization
In situ hybridization with digoxigenin-labelled RNA
probes was carried out according to the procedure established by Schulte-Merker (1993), but without acetylation
and RNase steps. Embryos were dechorionated with forceps and further fixed in 4% PFA in PBS at 4°C overnight. Embryos were treated with proteinase K for periods
ranging from 30 s to 15 min depending on the stage. Embryos were prehybridized for 2 h, hybridized overnight
with l-3 @ml probes and washed at 58°C. The hybridization buffer contained 5 mg/ml torula RNA (Fluka). The
alkaline phosphatase-conjugated anti-digoxigenin antibody (Boehringer) was preadsorbed (1:400 dilution) with
6 mg/ml adult zebrafish acetone powder (StrUe and Ingham, 1993). Incubation was for 2-4 h in PBST (PBS plus
0.1% Tween-20) containing 2 mg/ml BSA and 2% goat
serum. The antibody was diluted 1:2000. The alkaline
phosphatase reaction was allowed to develop for 1 h to
overnight depending on the age of the embryos. Embryos
were mounted in glycerol for observation. The slides
were cleared in xylene and prepared in Balsam of Canada.
Pictures were taken with an Axiophot photomicroscope
(Zeiss) and Nomarsky optic using Kodak Ektachrome 320
T films. After whole mount hybridization, some specimens were sectioned as described by Str;ihle and Ingham
(1993) in order to allow a precise histological analysis.
Acknowledgements
We would like to thank Dr. C. Fromental-Ramain for
the fish cDNA library, Drs. T. Gerster, G. Hauptmann and
F. Rosa for advice and fish strains as well as Dr. G.
Kraemer, members of the laboratory for helpful comments and discussions and Dr. E Barcellona for his help
with ethymology. This work was supported by funds from
the Swiss National Research Foundation, the Canton de
Genbve, the Claraz Foundation, the Latsis Foundation and
the HFSPO.
References
Brown, R. and Brockes, J.P. (1991) Development 111,489496.
Burke, A.C., Nelson, C.E., Morgan, B.A. and Tabin, C.J. (1995) Development 121,333-346.
Celnicker, S.R., Sharma, S. Keeian, D.J. and Lewis, E.B. (1990) EMBO
J. 9.42774286.
Chisholm, A. (1991) Development 111,921-932.
F. van der Hoeven et al. I Mechanisms of Development 54 (1996) 9-21
Condie, B.G and Cape&i, M.R. (1994) Nature 370.304-307.
Davis, A.P. and Capecchi, M.R. (1994) Development 120,2187-2199.
Dolle, P., Izpistla-Belmonte, J.-C., Falkenstein, H., Renucci, A. and
Duboule, D. (1989) Nature 342,767-772.
Doll& P, Izpisua-Belmonte, J.-C., Boncinelli, E. and Duboule, D.
(1991) Mech. Dev. 36,3-13.
Doll& P, Brown, J.M., Tickle, C. and Duboule, D. (1991) Genes Dev.
5, 1767-1776.
Doll& P., Dierich, A., LeMeur, M., Schimmang, T., Schuhbaur, B.,
Chambon, P. and Duboule, D. (1993), Cell 754314ll.
Doll& P., Fraulob, V. and Duboule, D. (1994) Development (Suppl.),
143-153.
Duboule, D. (1992) BioEssays 14, 375-384.
Duboule, D. (1994) Development (Suppl.), 135-142.
Duboule, D. and Doll& P. (1989) EMBO J. 8, 1497-1505.
Ericson, J.U., Krauss, S. and Fjose, A. (1993) Int. J. Dev. Biol. 37,
263-272.
Fjose, A., Izpistla-Belmonte, J.-C., Fromental-Ramain, C. and Duboule,
D. ( 1994) Development 120.7 l-81.
Garcia-Fernandez, J. and Holland, P.W.H. (1994) Nature 370, 563-566.
Gaunt, S.J., Sharpe, P.T. and Duboule, D. (1988) Development (Suppl.)
104, 169-179.
Gaunt, S.J. and Strachan, L. (1994) Dev. Dynam. 199,229-240.
Goodrich, E.S. (1913) Q. J. Micmsc. Sci. 59,227-248.
Graham, A., Papalopulu, N. and Krumlauf, R. (1989) Cell 57,367-378.
Hinchliffe, J.R. (1991) In Hinchliffe, J.R., Hurle, J.M. and Summerbell,
D. (eds.), Developmental Patterning of the Vertebrate Limb, Plenum Press, New York. pp. 313-323.
Holland, P.W.H., Garcia-Femandez, J., Williams, N.A. and Sidow, A.
(1994) Development (Suppl.), 125-133.
Holland, P.W.H. (1991) Gene 98.253-257
Izpisua-Belmonte, J.-C., Tickle, C., Dolle, P., Wolpert, L. and Duboule,
21
D. (1991) Nature 350,585-589.
Joly, J.-S., Joly, C., Schulte-Merker, S., Boulekbache, H. and Condamine, H. (1993) Development 119, 1261-1275.
Kostic, D. and Capecchi, M.R. (1994) Mech. Dev. 46.23 l-247.
Kmmlauf, R. (1994) Cell 78, 191-201.
Le Mouellic, H., Lallemand, Y. and Briilet, P. (1992) Cell 69,251-264.
Lewis, E.B. (1978) Nature 276.565-570.
McGinnis, W. and Krumlauf, R. (1992) Cell 68,283-302.
Njolstad, P.R., Molven, A., Apold, J. and Fjose, A. (1990) EMBO J. 9,
515-524.
Pendleton, J.W., Nagai, B.K., Murtha, M.T. and Ruddle, F.H. (1993)
Proc. Natl. Acad. Sci. USA 90,6300-6304.
Ramirez-Solis, R., Zheng, H., Whiting, J., Krumlauf, R. and Bradley,
A. (1993) Cell 73.279-294.
Rijli, F.M., Mark, M., Lakkaraju, S., Dierich, A., Doll&, P. and Chambon, P. (1993) Cell 75, 1333-1349.
Roberta, D.J., Johnson, R.L., Burke, A.C., Nelson, CE, Morgan, B.A.
and Tabin, C (1995) Development, in press.
Schulte-Merker, S. (1993) In Westertield, M. (ed.), The Zebraflsh
Book, University of Oregon Press. pp. 9.16-9.21.
Stmhle, U. and Ingham, P.W. (1993) Zebrafish Science Monitor 2,5-7.
Streisinger, G., Singer, F., Walker, C., Knauber, D. and Dower, N.
(1986) Genetics 112,311-319.
Sordino, P., van der Hoeven, F. and Duboule, D. (1995) Nature 375,
678-681.
Tucker, AS. and Slack, J.M.W. (1995) Development, 121.249-262.
Wang, B.B., Miiller-Immergltick, M. M., Austin, J., Robinson, N.T.,
Chisholm, A. and Kenyon, C. (1993) Cell 74, 29-42.
Westetfield, M. (1993) The Zebrafish Book. University of Oregon
FJRSS.
Yokouchi, Y., Sakiyama, J.-I. and Kumiwa, A. (1995) Dev. Biol. 169,
76-89.