Introduction

The Hox complex was originally characterised in Drosophila melanogaster (Lewis 1978; Kaufman et al. 1990). It contains eight genes with roles in anteroposterior (AP) patterning, which have all been shown to be related and to contain a homeobox motif. The molecular characterisation of the Hox complex revealed further genes containing the homeobox motif, which yet lack Hox-like expression or a homeotic function. These genes include bicoid (bcd), zerknüllt (zen; and the closely related zen2 [z2]) and fushi tarazu (ftz). Comparisons of their amino acid sequences with those of Hox proteins from other animals have shown that bcd, zen and z2 are closely related to each other (coming from duplications in the dipteran lineage) but are all orthologous to the Hox3 genes of non-insect animals (Falciani et al. 1996) and that ftz is most likely an ortholog of the protostomian Lox5 gene (Telford 2000). Importantly, both Hox3 and ftz/Lox5 genes from non-insects arthropods have the restricted AP expression pattern typical of Hox genes suggesting the homeotic function has been conserved (Abzhanov et al. 1999; Damen and Tautz 1998; Telford and Thomas 1998; Telford 2000; Hughes and Kaufman 2002a; Damen et al. 2005; Janssen and Damen 2006).

In insects, the Hox3 ortholog shows a complex history of gene duplication as well as changes of function (see Panfilio et al. 2006), but the main step in its evolution of a novel function seems to involve a transition from a Hox-like role in AP patterning with discrete anterior and posterior boundaries to an involvement in patterning the extra-embryonic tissues (Hughes et al. 2004). The timing of this change of function has not been precisely determined as no Hox3 expression data are yet available from a crustacean.

In Drosophila melanogaster, ftz has two functions: an early role in segmentation as a pair-rule gene (expressed in a series of seven stripes) and a later involvement in the specification of certain neurons (Wakimoto et al. 1984; Doe et al. 1998). The absence of homeotic function has been correlated with the loss of the hexapeptide motif (YPWM) upstream of the homeodomain (which is involved in the interaction with the cofactor Extradenticle, important for the homeotic function of Hox genes) and the acquisition of a LXXLL motif in the insect lineage, which is involved in the pair-rule function (Löhr and Pick 2005). It is unclear to what extent the pair-rule function and/or the role in segmentation are conserved within the insects, although the neurogenic expression is consistently observed (Stuart et al. 1991; Brown et al. 1994; Dawes et al. 1994; Hughes et al. 2004). In myriapods (millipedes and centipedes) and chelicerates (spider), ftz displays both a neurogenic and a Hox-like pattern of expression (respectively, Janssen and Damen 2006; Hughes and Kaufman 2002a; Damen et al. 2005). Although the Hox-like expression is observed, it is not known if ftz is expressed in the central nervous system (CNS) of the mite Archegozetes (Chelicerata, Telford 2000). Additionally, involvement in segmentation has been suggested in the case of Lithobius (Myriapoda, centipedes, Hughes and Kaufman 2002a). ftz has been studied so far in only one crustacean, the morphologically derived barnacle Sacculina carcini (cirripede), in which it is apparently restricted to the CNS (Mouchel-vielh et al. 2002). The current view, then, is that the homeotic function of ftz has been lost at an unknown stage in the lineage leading to the insects, probably associated with the gain and loss of some important cofactor interaction motifs (Alonso et al. 2001; Löhr et al. 2001; Hughes et al. 2004; Löhr and Pick 2005).

One important challenge for evolutionary biologists is to discover the genetic changes resulting in novel gene functions and to understand how a gene with an important role in development can lose that role without disastrously disrupting embryogenesis. Hox3 and ftz are members of the Hox gene family, which have drastically changed their role in arthropods and thus provide an opportunity to study the mechanisms underlying the functional evolution of Hox proteins.

We have studied the expression of Hox3 and ftz in the crustacean Daphnia pulex (Branchiopoda). This model organism presents several advantages being easy to breed, producing a fairly large number of eggs and having an almost completely sequenced genome.

In the following discussion, we use our expression data to inform our model of Hox3 and ftz evolution, but it is important to remember that the model would benefit from additional functional data.

Materials and methods

Husbandry

Daphnia pulex specimens were kindly provided by Dr John Colbourne of the Daphnia Genomics Consortium (http://daphnia.cgb.indiana.edu/). Animals are kept in a mixture of tap water and distilled water and fed with the algae Scenedesmus acutus, bought from the University of Toronto Transcript Center http://www.botany.utoronto.ca/utcc/.

Gene isolation and WMISH

The first draft of the Daphnia pulex genome (provided by DoE Joint Genome Institute & the Daphnia Genomics Consortium, http://daphnia.cgb.indiana.edu/) was used as a target for TBLASTX using arthropod Hox proteins as query sequences to retrieve orthologs of Hox3 and ftz. Probes for Whole Mount In Situ Hybridisation (WMISH) were amplified using polymerase chain reaction (PCR) primers upstream of the homeobox to avoid non-specific signal because of homeobox sequence conservation. Primers used for Hox3 were ACTGCTCGTCTGTTGATGTCGG (forward) and CGTTTGCTGGTTGTTCAGG (reverse) and for ftz GCTTCCAGCTGTACCTCACC (forward) and GTAGGCTTGCATCCAGCTTC (reverse). Amplified complementary DNAs were cloned into the pGEM-T easy vector (Promega). Accession numbers: EF363483 (DpuHox3), EF363484 (Dpuftz).

The protocol used for WMISH was as described by Shiga et al. (2002), with the following modifications: vitelline membranes were removed by peeling, no sonication step and antibody washes were overnight at 4°C. DIG-labelled probes were synthesised using the Roche DIG-labelling KIT. At the end of the WMISH, embryonic nuclei were fluorescently labelled using 4′-6-diamidino-2-phenylindole (DAPI).

Embryos were recorded using the ZEISS Imager M1 microscope equipped with a ZEISS AxioCam HRc and the program AxiVision 4.3. For some embryos (Fig. 2a–d) series of pictures at different focal planes were taken and merged using the stacks/Z-project option of ImageJ (Rasband, W.S., ImageJ, US National Institutes of Health, Bethesda, MD, http://rsb.info.nih.gov/ij/, 1997–2006) so that the entire expression pattern could be seen on a single picture.

Embryos were staged according to the study of late Daphnia embryogenesis by Kotov and Boikova (2001). Fifty hours after the beginning of development, the juvenile, which is essentially a miniature adult, is released from the mother’s brood pouch.

Results and discussion

Expression of DpuHox3

We found only one copy of the Hox complex in Daphnia, and the Hox3 sequence was PCR amplified and cloned as described. Comparison of the homeodomain sequence with those from other species demonstrates the high conservation typical of most Hox proteins (Fig. 1). The hexapeptide YPWM is present upstream of the homeodomain of DpuHox3 (data not shown), whereas the various zen-specific motifs proposed in Falciani et al. (1996) and Stauber et al. (2002) could not be found.

Fig. 1
figure 1

Daphnia pulex Hox3 (upper panel) and Ftz (lower panel) homeodomain amino acid sequences compared with orthologs from other arthropods. Conserved residues are shaded and unknown residues are left blank. In the lower panel, the presence or absence (dashes) of the two cofactor-interactions motifs LXXLL and the hexapeptide (Hexap) are shown (see the text for discussion). Numbers indicate the length of the sequences separating the domains from each other and the hexapeptide from the homeodomain. The Daphnia homeodomain sequences were obtained from the genome data available on https://daphnia.cgb.indiana.edu. Abbreviations: Hexap Hexapeptide, C Crustacea, Ch Chelicerata, I Insecta, M Myriapoda, O Onychophora, T Tardigrada. Species used: Aga Anopheles gambiae, Aka Acanthokara kaputensis, Alo Archegozetes longisetosus, Ate Achaearanea tepidorium, Cma Carcinus maenas, Csa Cupiennius salei, Dme Drosophila melanogaster, Dpu Daphnia pulex, Gma Glomeris marginata, Lat Lithobius atkinsoni, Mta Milnesium tardigradum, Ofa Oncopeltus fasciatus, Sge Schistocerca gregaria, Tca Tribolium castaneum, Tdo Thermobia domestica, Uoe Ulophysema oeresundense. Asterisk in DmeFtz, the length corresponds to the number of positions between the LXXLL motif and the homeodomain as there is no hexapeptide

Full size image

WMISH with the Daphnia pulex Hox3 gene (DpuHox3) revealed a dynamic expression profile (Fig. 2). In early embryos (12 h, Fig. 2a–c), only the second antennal grooves are visible. The first antennal grooves are very small, and the other appendages have not yet started to form, making it difficult to determine in which segments the gene is expressed. These embryos show a complex pattern of expression, with three symmetrical regions of expression, a dorsal band and an additional site of expression on the future location of the posterior furrow (Fig. 2a–c). At 14 h, when the mandibles enlarge, only the middle pair of cell groups still noticeably express DpuHox3 (Fig. 2d). The other pairs of expression have nearly completely faded.

Fig. 2
figure 2

Hox3 expression in Daphnia pulex. A–C, 12-h embryo. A, ventral view. The first antennal grooves are very small, and the other appendages have not yet started to form, making it difficult to determine in which segments the gene is expressed. Three symmetrical pairs of region expression are observed: the most anterior pair is probably on the anterior part of the second antennae (1), the middle pair is more central and probably in the posterio-proximal part of the second antenna at the intersection with the antero-proximal region of the mandibles (2), the most posterior region of expression may be at the posterior limit of the future mandibles (3). B, dorsal view of the same embryo as A. There is a stripe of expression at the level of the second antennae and an additional region of expression corresponding to the future posterior furrow (Pf). C, left-side view of a different 12-h embryo. Ventral is to the left, dorsal to the right. When the dorsal band reaches the antero-distal part of the second antenna, it appears to divide. One band follows the anterior limit of the second antenna (1 in A), while the other one goes around the distal part of the same appendage. This latter band divides again into two, one following the posterior limit of the second antenna (eventually becoming stronger at the end, which corresponds to the group of cells 2 in A), and the other one corresponds to pair of stripes 3 in A. D, ventral view of a 14-h embryo. The mandibles are forming, and only the middle groups of cells are still noticeably expressing DpuHox3. The other paired regions of expression observed earlier have nearly completely faded while the dorsal domains of expression have not changed (data not shown). E–F, expression can only be detected in the limb buds of the mandibular segment in 15-h (E), 16-h (F), 17.5-h (G) and 20-h (H) embryos. Absence of mRNAs in the outer layer of the mandible buds (arrowhead in E′, F, F′, G and G′) suggests that DpuHox3 is only expressed in the mandibular mesoderm. H–J, series of focal planes from a 20-h embryo. Comparisons to the same series with DAPI counterstaining (H′–J′) show that the DpuHox3 mRNAs are not localised in the external layer of the embryo, supporting an expression in the mesoderm of the mandible buds. A′–G′ DAPI counterstaining of the embryos shown in A–G. Abbreviations; A2 second antenna, Lr labrum, Md mandibles, Mx1 first maxilla, Mx2 second maxilla, Mxz maxillary zone, Pf posterior furrow, TS trunk segments. In all embryos, anterior is up. Scale represents 100 μm in A–D and H–J, 50 μm in E–G

Full size image

Around 15 h after the beginning of development (Fig. 2e), the pattern of DpuHox3 expression includes a domain strictly restricted to the mouthparts, in the mandible buds, but not in the ventral ectoderm of the corresponding segment. This expression in the mandibles is conserved until at least 20 h of development (Fig. 2e–h).

As the peripheral layer of the mandible buds is free of transcripts (arrowheads in Fig. 2e′, f, f′, g and g′), it seems that DpuHox3 is only expressed in the mesodermal layer of the limb. This is confirmed by series of pictures taken in parallel both with transmitted light and DAPI fluorescence; when the focal plane is on the external part of the embryo, the cells expressing Hox3 are not in focus (Fig. 2h, h′), but when the focal plane is deeper within the embryo, external features such as appendage buds become blurry, while the cells expressing DpuHox3 come into focus (Fig. 2i, i′ ,j and j′).

When compared to the Hox3 expression described in other arthropods, the pattern in the present study raises interesting points.

First, it is noteworthy that the Hox-like DpuHox3 expression is very restricted along the AP axis, ultimately being present only in the mandibles. As has happened with a number of Hox genes, particularly in the anterior region in insects, crustaceans and myriapods, the broad, overlapping pattern of expression seen in chelicerates is reduced and Hox expression is more restricted along the AP axis. This shrinking tendency may be linked to the diversification of feeding appendages in Mandibulata, notably in crustaceans and insects (Hughes and Kaufman 2002b).

Second, Daphnia is not the only arthropod in which a Hox3 gene is expressed in a ring (the basal Thysanura insect Thermobia and the millipede myriapod Glomeris) and dorsally in Glomeris (Hughes et al. 2004; Janssen and Damen 2006).

Although the posterior furrow of Daphnia and the posterior growth zone of other arthropods cannot be homologised, Hox3 is similarly expressed in a posterior region in Daphnia, Glomeris and Thermobia. The Hox3 expression in annelids also shares some of the above features such as a ring-shaped expression (Chaetopterus) or at least some dorsal domains (Nereis and Platynereis), as well as transcripts in the posterior prepygidial zone (Chaetopterus and Nereis; Irvine and Martindale 2000; Kulakova et al. 2007).

Moreover, the most anterior limit of DpuHox3 seems to be the anterior boundary of the second antenna, which corresponds to the most anterior limit of Hox3 in chelicerates (pedipalps) and myriapods (premandibular or intercalary segment) (Abzhanov et al. 1999; Damen and Tautz 1998; Telford and Thomas 1998; Hughes and Kaufman 2002a; Janssen and Damen 2006). This strongly suggests that this anterior boundary is ancestral. The symmetrical patch of expression at the anterior limit of the second antenna in Daphnia is also very similar to the early expression in the premandibular segment of Lithobius (Hughes and Kaufman 2002a).

The mandibular expression of Hox3 in Lithobius, Thermobia and Daphnia shows an additional level of similarity, in that transcripts are expressed only in the presumptive mesoderm of the limb bud and not ectodermally as with most Hox genes (Hughes and Kaufman 2002a; Hughes et al. 2004). This could represent a shared and derived pattern of expression associated with the origin of the mandibles—the defining character of the mandibulate clade.

Finally, in terms of the evolution of Hox3 in arthropods, the results presented here are in accordance with the model of evolution proposed by Hughes et al. (2004), with the conservation of the Hox-like expression in most arthropods, and its loss, together with the appearance of the extra-embryonic expression, in the lineage leading to Neoptera after the divergence of Thysanura. The Hox-like expression can be correlated with the presence of the hexapeptide motif upstream the homeodomain in Daphnia and the spider Cupiennius (as well as in Thermobia [Panfilio and Akam 2007]; it is not known if the hexapeptide is also present in the mite and the myriapods Glomeris and Lithobius, as the sequences available lack this portion of the protein) and other features that differentiate Hox-like Hox3 genes from zen-like ones (in higher insects) such as a more C-terminal position of the homeodomain in the protein and a longer protein length (Panfilio and Akam 2007).

Expression of Dpuftz

The homeodomain of Dpuftz is very similar to those of other arthropods sequences, however, the full length protein presents two interesting features. First, the hexapeptide is very derived (Fig. 1), even compared to the other available crustacean sequence (Sacculina). It is noteworthy that no hexapeptide could be found in Glomeris Ftz. In addition, the putative cofactor-interaction motif LXXLL (LNPLL in Daphnia) is present upstream of the homeobox at roughly the same position as in the Drosophila and Tribolium proteins. This is surprising as this motif is involved in the interaction of Ftz with its cofactor Ftz-F1 (which has been shown to be important for the pair-rule function of Ftz in insects) and is not present in Schistocerca gregaria (Alonso et al. 2001). Whether this motif has been lost in Schistocerca or there have been independent acquisitions in higher insects and Branchiopoda is uncertain. Although the motif lies at the expected position in the protein, and it cannot be found anywhere else in any other arthropod Ftz sequence, it should be stressed that the consensus sequence of this motif (LXXLL) is not very complex, and random substitution could explain its presence in Daphnia Ftz protein.

Like DpuHox3, Dpuftz has a dynamic expression (Fig. 3). Early on (13-h embryos, Fig. 3a–d), the transcripts form a ring around the antero-dorsal part of the embryo, at a similar position than DpuHox3. This ring then follows, on the ventral side, the posterior boundary of the mandibular segment (anterior to the future maxillary zone). At 14 h of development (Fig. 3e), Dpuftz transcripts accumulate in a restricted part of the earlier pattern, on the ventral midline, between the mandibles and the first maxillae, while the circum-embryonic part of the expression starts to fade, particularly on the ventrolateral parts of the embryo. Comparison with the expression of the segmental marker gene engrailed (Fig. 3h) reveals that in later stages (15.5-, 16- and 17-h embryos, Fig. 3f–g, i and j, respectively), Dpuftz is restricted to the posterior part of the mandibular segment in the ventral ectoderm. It is not known whether this position corresponds to the first ventral ganglion as the formation of the CNS in Daphnia pulex is poorly known, hence, there is no direct evidence of a neurogenic expression for Dpuftz.

Fig. 3
figure 3

ftz expression in Daphnia pulex. A–D, 13-h embryos. A and B, ventral and dorsal sides of the same embryo. C and D, left and right sides of the same embryo. On the ventral side, the ring-shaped expression follows the posterior boundary of the mandibular segment (anterior to the future maxillary zone) and extends dorsally at the level of the distal limit of the second antenna. On the dorsal side, the band is similar to that observed with DpuHox3. E, 14-h embryo, ventral side. Dpuftz transcripts accumulate in the middle region of the ventral expression (arrowhead). The ring around the embryo starts to fade, particularly ventrolaterally. F, 15.5-h embryo. Strong expression can be seen only in a central region (arrowhead) in the posterior part of the mandibular segment, probably in the ventral ectoderm. G, I, J, close up on the region of Dpuftz expression in 15.5-, 16- and 17-h embryos. H, expression of Daphnia pulex engrailed mRNAs in the same region, showing that Dpuftz is expressed in the posterior part of the mandibular segment (asterisks). A′–J′, DAPI counterstaining of the embryos shown in A–J. Abbreviations: Cr Carapace. All other abbreviations are the same as in Fig. 2. In all embryos, anterior is up. Scale represents 100 μm in A–F, 50 μm in G–J

Full size image

In Daphnia, as mentioned above about DpuHox3, we found a much more restricted domain of ftz expression than is seen in chelicerates (Fig. 3; Telford 2000; Damen et al. 2005). It is noteworthy that the expression of ftz in Daphnia is completely different to that seen in the crustacean Sacculina, where there is no Hox-like pattern, only expression in reiterated symmetrical pairs of spots located in the head and trunk ectoderm, in what are thought to be neural cells (Mouchel-vielh et al. 2002).

As with DpuHox3, some possible conserved features have been observed between Dpuftz and its ortholog in annelids (Lox5) such as a dorsal and ring-shaped expression pattern (Kulakova et al. 2007).

The pattern of Dpuftz described here has several implications regarding ftz evolution in arthropods. First, it seems that the Hox-like expression (and, by extrapolation, the homeotic function) has been retained in Daphnia. It is now accepted that the insects derive from a paraphyletic assemblage of crustaceans, and the possible positioning of Branchiopoda as a close sister group to the Hexapoda (insects and relatives; Mallatt and Giribet 2006; Glenner et al. 2006) allows an even more precise pinpointing of this loss of “Hox-like behaviour”: probably in the lineage between the Branchiopoda and the basal Thysanura (Hughes et al. 2004). Given the present data, it can be proposed that the situation in Daphnia represents an intermediate evolutionary stage of ftz loosing its typical Hox function. Although this should be further investigated with functional studies, this conclusion is supported by the derived sequence of the hexapeptide in Dpuftz. The unexpected presence of the LXXLL motif upstream of the homeodomain (Fig. 1), which is thought to be important for the segmentation function, is quite puzzling, as there is no evidence of a role in segmentation for ftz in Daphnia.

Second, if the expression pattern of ftz in the myriapod Lithobius is indeed linked to a segmentation function, as claimed by the authors (Hughes and Kaufman 2002a), there are two evolutionary scenarios. If this role is homologous to the putative segmentation function in basal insects (a function different to the pair-rule function, which is specific to higher insects, and probably to Drosophilids), then one would be forced to conclude that this segmentation function has been lost in crustaceans and in millipedes (this study; Mouchel-vielh et al. 2002; Janssen and Damen 2006). Alternatively, the segmentation-like pattern and the proposed segmentation function in myriapods could be convergent on that in insects. We favour this last hypothesis. At this point, it is also interesting to mention first, that it is postulated that the expression pattern observed in Schistocerca is not related to a segmentation function (Dawes et al. 1994) and second, deletion or RNA interference knock down of Tribolium ftz fail to cause a segmentation defect (Stuart et al. 1991; Choe et al. 2006); these are the only functional assays of ftz outside Drosophila.

Our examination of ftz and Hox3 in Daphnia reveals characteristics of expression that show differing degrees of phylogenetic distribution and hence conservation. First, an aspect of expression likely peculiar to the water fleas is the continuous stripe of both DpuHox3 and Dpuftz expression around the circumference of early embryos; this is not a general feature of arthropods or even crustaceans and presumably depends on the morphogenetic movements involved in the formation of the embryonic blastoderm. Second, we find that expression of Hox3 in the mandibles of all three groups of mandibulate arthropods (insects, crustaceans and myriapods) is ultimately restricted to the mesoderm; a detail that differs from the chelicerates and might give some support to the homology of mandibles across these taxa. Finally, the anteriormost boundary of DpuHox3 expression corresponds with that seen in myriapods and chelicerates—the second antennal/premandibular/pedipalpal segment. Although our results suggest that these two Daphnia Hox genes have retained their ancestral function in AP patterning when compared to the novel roles seen in insects, it is nevertheless clear from these observations that their use as Hox genes has been subject to numerous more or less subtle alterations.