Homeobox genes in the ribbonworm Lineus sanguineus: Evolutionary implications

Abstract

From our current understanding of the genetic basis of development and pattern formation in Drosophila and vertebrates it is commonly thought that clusters of Hox genes sculpt the morphology of animals in specific body regions. Based on Hox gene conservation throughout the animal kingdom it is proposed that these genes and their role in pattern formation evolved early during the evolution of metazoans. Knowledge of the history of Hox genes will lead to a better understanding of the role of Hox genes in the evolution of animal body plans. To infer Hox gene evolution, reliable data on lower chordates and invertebrates are crucial. Among the lower triploblasts, the body plan of the ribbonworm Lineus (nemertini) appears to be close to the common ancestral condition of protostomes and deuterostomes. In this paper we present the isolation and identification of Hox genes in Lineus sanguineus. We find that the Lineus genome contains a single cluster of at least six Hox genes: two anterior-class genes, three middle-class genes, and one posterior-class gene. Each of the genes can be definitely assigned to an ortholog group on the basis of its homeobox and its flanking sequences. The most closely related homeodomain sequences are invariably found among the mouse or Amphioxus orthologs, rather than Drosophila and other invertebrates. This suggests that the ribbonworms have diverged relatively little from the last common ancestors of protostomes and deuterostomes, the urbilateria.

The genetic basis of development of vertebrates and arthropods has been extensively studied during the last decade. It has become clear that although these organisms have strikingly different body architectures, many of the regulatory genes they use to establish their body plan are conserved (1). Most prominent among them are the homeobox genes that encode gene regulatory proteins containing a DNA-binding homeodomain. Initially identified in Drosophila, they are present in many, if not all, eukaryotic organisms (2). Hox genes, a subfamily of homeobox genes, are involved in the specification of body structures along the anterior–posterior (A-P) body axis in all animals analyzed so far, including arthropods and chordates (3–6). In arthropods, nematodes, and chordates, these genes are organized in clusters (7). During development, their physical linear order is reflected in their respective region-specific expression domains along the A-P axis (8). Thus, it has been suggested that the physical linear order of the ancestral Hox complex is mechanistically linked to the spatial and temporal order of their expression. This is referred to as spatiotemporal colinearity (9). It has been proposed that different body patterns may evolve through changes in number, regulation, or function of Hox genes (10). Knowledge of the evolutionary history of Hox genes will allow us to compare different body plans and thus will provide insights into how new body plans have evolved. Until recently, extensive studies of Hox genes have been performed mainly on flies and vertebrates. To enable us to draw a reliable scenario about how Hox genes have evolved, studies of Hox genes have to be broadened to include representatives of other phyla, in particular different protostomes.

Based on the original hypothesis of Lewis (11) that the Bithorax-Complex of Drosophila arose by unequal crossing-over, leading to tandem duplications of the Hox genes, we have proposed that the primordial Hox cluster may have arisen by a series of consecutive tandem duplications from unequal crossing-over (12). The first step is thought to have led from the original prototype gene (Ur-Hox gene) to the two terminal genes of the cluster, which evolved into an anterior and posterior gene, respectively, and, therefore, had the longest time to diverge. Subsequently a series of unequal crossings-over presumably generated the interior genes. Because of these multiple recombination events, the internal genes are mosaics, combining segments of their ancestral genes, and their sequences become homogenized. Because the interior genes were generated later in evolution and were homogenized, they have diverged less from the Ur-Hox gene than the external genes. This hypothesis is consistent with the finding that the homeodomain sequences diverge progressively from the consensus sequence, going from the center toward the termini (12). It is also supported by the fact that the terminal genes share an intron at the same position in the homeodomain (amino acids 44/45) (12). The progressive divergence of the homeobox sequences from the center toward the termini of the clusters is also apparent in a distance matrix based on pairwise sequence comparisons in all possible combinations to quantify sequence similarities (13), as well as in phylogenetic trees that are rooted on the Antennapedia (Antp) gene, that corresponds most closely to the consensus sequence and may have conserved the sequence of the Ur-Hox gene best (12). A quantitative phylogenetic analysis using the neighbor-joining method basically supports the above model (14). The phylogenetic tree indicates that the Ur-Hox gene first duplicated and diverged into an anterior and a posterior gene. In vertebrate phylogeny, the anterior gene, through a series of duplications, gave rise to eight anterior cognate group genes, whereas the posterior gene duplicated repeatedly and diverged into five posterior group genes. The sequence and time of these duplications can be roughly estimated (14). To test the predictions of the unequal crossing-over model and the phylogenetic distance tree, we have to obtain information about evolutionary intermediates (if they still exist) that did not undergo all of the gene duplication events.

Lineus, a marine ribbonworm, may represent such an intermediary stage of Hox gene evolution. The phylogenetic position of the ribbonworms (nemertini) has been much debated (15–19). They belong to the triploblasts. Lineus is bilaterally symmetrical and has a rather sophisticated organization along the anterior–posterior (A-P) body axis. These different A-P body regions are defined according to specific morphological characteristics such as eyes in region 1 (R1), brain in R2, mouth in R5, and anus in R9 (Fig. 1). Nemerteans have a lateral nervous system, rather than a ventral one as found in most other invertebrate phyla. They share spiral cleavage with platyhelminths, annelids, and mollusks. They have a body cavity with an eversible proboscis (rhynchocoel) and blood vessels that have been interpreted as coelomic cavities (20). In earlier studies, it has been suggested that living nemerteans might be similar to the last common ancestor between invertebrates and vertebrates (16–19). These last common ancestors have been termed Urbilateria and presumably represent primitive coelomates (21). Animals possessing a coelomic cavity have been subdivided into two fundamental groups: the protostomes, such as arthropods, annelids, and mollusks, and the deuterostomes, such as echinoderms and chordates (chordates include ascidians, Amphioxus, and vertebrates). In protostomes the blastopore gives rise to mouth, whereas in deuterostomes the mouth is formed secondarily by a perforation of the ectoderm, and the anus arises at or close to the site of the original blastopore (22). However, the embryological distinction between protostomes and deuterostomes is not as clear as it would seem, especially in the case of nemerteans. In polychaetes the blastopore is elongated along the anteroposterior axis and the lateral blastopore lips fuse along the midline, leaving two openings, an anterior one, which will form the mouth, and a posterior one giving rise to the anus (23). The snail Viviparus and the polychaete Eunice show deuterostomy (23), even though they are classified as protostomes.

Figure 1.

Scheme of Lineus body organization. The different body regions (R) along the A-P axis are numbered starting from the anterior end and are characterized by the following structures: R1, cephalic glands, rhynchodeum, and eyes; R2, cerebral ganglia; R3, sensory cerebral organs; R4, postcerebral, preesophageal connective tissue; R5, mouth; R6, posterior esophagus and nephridia; R7, anterior intestine (gonads absent); R8, middle and posterior intestine plus gonads; R9, anus; R10, caudal end.

Embryological studies on Lineus ruber indicate that after gastrulation the blastopore closes almost completely and that the mouth forms as a secondary invagination of the ectoderm, close to the site of the closed blastopore (24). Furthermore, the axis of the larva (and adult) relative to the polarity of the egg may change as a function of foregut morphogenesis in the embryo (25). During foregut formation the mouth moves to various degrees from the vegetal pole to the ventral side of the body, e.g., in Lineus torquatus the stomodaeum moves toward the opposite side of the ptilidium larva (25). This suggests that nemerteans may be in an evolutionary transition zone between proto- and deuterostomes, a point that is worth reexamination. The sequence data of 18S ribosomal RNA give a relatively clear picture; the nemertean Cerebratulus lacteus falls within the protostome coelomate clade (26). These data have been confirmed and extended by Carranza et al. (27) by comparing the complete 18S rRNA sequences of 35 species from platyhelminths and various other phyla, including the nemertean Prostoma eilhardi, which is most closely related to annelids and mollusks within the protostomes and more closely related to deuterostomes (Fig. 2) than to arthropods and platyhelminths. In a similar study of 18S rRNA sequences the nemertean Lineus sp. was also found to be closely related to mollusks and annelids (28).

Figure 2.

Evolutionary distance tree. Based on the neighbor-joining method including the complete 18S ribosomal RNA sequences of 33 species belonging to 16 different phyla, with special emphasis on platyhelminths. The nemertean listed is Prostoma eilhardi. The numbers at the nodes are percentages of 1,000 boot trap replicates that support the branch, with only values over 50% being represented. All branch lengths are drawn to scale. The choanoflagellate Sphaeroeca volvox is used as outgroup. The Fitch–Margoliash method gives the same topology. (Reproduced with permission from ref. 27.)

We have analyzed the Hox genes of Lineus sanguineus and found them to be clustered in a relatively small segment of DNA that can be resolved by pulse-field electrophoresis. The cluster contains at least six Hox genes, corresponding to the vertebrate homologs Hox1 and 3 (anterior class), Hox 4, 6, and 7 (middle class), and Hox 9 (posterior class). Because at least one member (Hox5) present in vertebrates and Amphioxus seems to be missing in Lineus, the Hox cluster of nemerteans may be an early evolutionary intermediate, unless the missing gene has been lost secondarily. Interestingly, the homeodomain sequences of Lineus are more closely related to those of vertebrates and Amphioxus, rather than to those of Drosophila and other invertebrates.

MATERIALS AND METHODS

DNA Extraction.

To isolate genomic DNA we followed the procedure described by Hempstead et al. (29) with few modifications. All steps were performed at 4°C, and the SDS concentration in the extraction buffer was increased to 5%.

Cloning Homeobox Fragments and Hox Genes from Genomic DNA.

Homeobox fragments were amplified from genomic DNA as described by Murtha and Ruddle (30) and PCR products were cloned into the Bluescript vector.

The L. sanguineus genomic library used is a λ Fix II library prepared with MboI partially digested DNA (31). High-stringency screening of the library was performed as described by Sambrook et al. (32). For the low stringency screens, hybridization temperatures were chosen to allow cross-hybridization between related homeobox sequences. Using a hybridization buffer containing 50% formamide, we reduced the hybridization temperature to 37°C.

Pulse-Field Gel Electrophoresis. Preparation of embedded DNA.

Worms were flash-frozen in liquid nitrogen and ground to fine powder with a liquid nitrogen-chilled mortar and pestle. The powder was immediately mixed to 1% low-melting agarose. The embedded worm powder was then treated with proteinase K (33). For endonuclease digestion, a liquid digestion procedure was used (33). Complete digestion was obtained after overnight incubation at 37°C whereas 2-hr digestion was performed for partial digestion.

Pulse-field gel electrophoresis conditions.

Gel (0.8%) was run with fields of 6 V/cm by using 120° reorientation and a switch interval of 1.5 sec. The run was performed during 12 hr at 14°C with 0.5× TAE running buffer (Tris/acetate/EDTA). We have set up these conditions for the Chef Mapper apparatus (Bio-Rad).

Alkaline capillary transfer was performed for 48 hr, and the hybridization procedure was the same as described above.

RESULTS AND DISCUSSION

Amplification of Homeobox Sequences by PCR.

To identify L. sanguineus homeobox genes, we have performed a PCR screen on genomic DNA. For this purpose we have used degenerate primers derived from the consensus coding sequences of the first and third α helices of the homeodomain (30). Analysis of the PCR products by agarose gel electrophoresis showed the presence of a single band. DNA from this band was subcloned. Of 50 clones sequenced, we found five different homeobox sequences. Based on both nucleotide and amino acid sequence analysis, one has been classified as a caudal-type homeobox and the four others as Hox-type homeoboxes. Sequence data obtained from these short DNA fragments were not sufficient to assign these four Hox-type homeoboxes to defined Hox-class genes. Therefore, we screened a genomic library to clone and analyze more coding sequences of these genes.

Isolation and Sequence Analysis of L. sanguineus Hox/HOM-Type Genes.

We used the isolated PCR fragments as probes to screen a L. sanguineus genomic library under high-stringency conditions. For each homeobox probe, we isolated multiple clones that were restriction-endonuclease mapped and partially sequenced. Sequences of these genes revealed that we isolated the corresponding gene for each homeobox probe, thus confirming the authenticity of all five homeoboxes amplified by PCR. To extend our screen to different Hox genes we have performed a second round of screening in which reduced stringency conditions were chosen to allow cross-hybridization between related Hox sequences. From this screen we have isolated 36 phages. Sequencing of phages with different restriction-endonuclease maps led to the identification of three new homeobox genes. Sequence analysis of the homeobox plus adjacent sequences allowed us to unambiguously classify each gene (the deduced amino acid sequences are shown in Fig. 3). Among the eight homeobox genes identified, six are Hox-type genes, one is a caudal-type gene (Lscdx), and one is a NK-1-type gene (LsNK). When the complete homeodomains with some flanking sequences are used for comparison, the Lineus sequences can be assigned unequivocally to their orthologs in vertebrates or Amphioxus. Therefore, we decided to discuss homology and assignment in terms of membership to an ortholog group according to the classification of vertebrate Hox genes into 13 paralog groups. Because each Lineus Hox gene can be assigned to a different ortholog group, we named these genes LsHox1, LsHox3, LsHox4, LsHox6, LsHox7, and LsHox9, according to the ortholog group to which they belong.

Figure 3.

Sequence alignments of homeodomains and flanking sequences encoded by LsHox genes (Upper) plus caudal/Cdx and NK-1 orthologs (Lower). Dashes indicate amino acid identity between domains encoded by L. sanguineus homeobox genes and the most similar sequences from other phyla. Percentages indicated correspond to sequence identities within the homeodomain between L. sanguineus genes and the sequences from other phyla. The sequence comparison is limited to the small number of homeodomains known in protostomes for which complete sequence data are available.

Sequence comparison with the known Hox genes gave an intriguing result because for each LsHox gene the highest sequence identity was found invariably with mouse and Amphioxus Hox genes. For each of the LsHox homeodomains (except LsHox9) the sequence identities with their chordate orthologs vary between 88 and 100%. One hundred percent of sequence identity was found between the LsHox7 and AmphiHox7 homeodomains. In the case of the LsHox9 homeodomain the highest percentage of sequence identity was found to be 69% only with the mouse Hoxc-9 homeodomain, but LsHox9 is clearly a posterior gene like Hoxc-9.

LsHox1 is assigned to ortholog group 1 based on its 90% sequence identity to Amphi Hox1 and 88% identity to mouse Hoxa1 and Drosophila labial. Among the 17 aa characteristic for the Hox1 group, 11 are shared by LsHox1. In the C-terminal flanking sequences 4 aa are shared with Hoxa1 and labial and 2 are shared with Amphi Hox1. In the N-terminal flanking sequences 3 aa are shared with Hoxa1.

LsHox3 shows 91% sequence identity with murine and human Hoxb3 and murine Hoxd3, whereas the Amphi Hox3 and Drosophila proposcipedia show 88 and 73% sequence identity, respectively. Ten of 13 characteristic amino acids are shared with the Hox3 group, whereas none of the 9 aa that are typical for Hox2 are found in LsHox3. In addition, two lysine residues are highly conserved at positions +1 and +3 of the C-terminal flanking sequences, which allows us to assign LsHox3 to ortholog group 3.

LsHox4 clearly belongs to ortholog group 4 because the 8 aa downstream to the homeodomain have 90% sequence identity with those of the mouse genes Hoxa4, Hoxb4, and Hoxd4 and also the upstream flanking sequences share 55% identity with Hoxa4, Hoxb4, Hoxc4, and Hoxd4. Again the closest similarity within the homeodomain is with a mammalian gene—90% with murine Hoxa4 and Hoxd4—whereas Drosophila Deformed and Amphi Hox4 show 88 and 86%, respectively.

LsHox6 is very similar to Amphi Hox6 (95% identity), but only 2 aa are shared in the flanking sequences. However, the 2 aa at positions 6 and 7 (Q and T) indicate that this gene belongs to ortholog group 6 and not to group 5, which has T and (S or A) at these positions. These amino acids have been shown to be functionally important in the distinction between Antp and Scr (34). LsHox6 shows 92% sequence identity and 5 of 10 identical amino acids in the C-terminal flanking sequences to the planarian Dthox-E.

LsHox7 can be assigned to ortholog group 7 because it shares 100% sequence identity with Amphi Hox7. There are also three characteristic amino acids in the N-terminal flanking and a lysin at position +1. This gene is more closely related to Antennapedia (98%) than to abdominal A (95%), and it lacks the sequences at the C terminus that encode the Ubd-A peptide (28). LsHox7 is more closely related to Antennapedia than LsHox6, especially in the N-terminal arm.

The only homeodomain that is more difficult to assign is LsHox9 because the maximum sequence identity shared with any known homeodomain is 68% for Hoxc9 and 67% for Hoxa9, Amphi Hox9, and Abdominal B of Drosophila. It lacks the Ubd-A peptide or other striking similarities in the flanking sequences. Nevertheless, the characteristic amino acids near the N terminus (T1, K3, P7, T12, and L21) together with the overall similarity identify LsHox9 clearly as a member of the posterior group.

The LsHox1 and LsHox7 also have a conserved motif in the C-terminal region, as do their respective vertebrate orthologs. LsHox1 shares a serine-rich domain with the vertebrate paralog group 1 genes, whereas LsHox7 has a glutamate stretch near the C terminus in common with the vertebrate paralog group 7 genes (Fig. 4) (34).

Figure 4.

Sequence alignments of the C-terminal domains encoded by the LsHox1 and LsHox7 genes with chordate OG-1 genes (Upper) and vertebrate PG-7 genes (Lower), respectively. Dashes indicate identical amino acids. Conservative changes and identical residues are boxed in.

Evolution of the Hox Gene Cluster.

To date, the genome organization of Hox genes has been investigated in humans, mouse, puffer fish, Amphioxus, Drosophila, Tribolium, and Caenorhabditis (35, 36). Amphioxus and the three invertebrate species have a single Hox cluster (37–40) whereas vertebrates possess multiple clusters, four copies in humans, mice (35), and puffer fish (36), and probably three in the most primitive jawless vertebrates (lampreys) (41), suggesting that several duplications of the entire cluster have occurred during vertebrate evolution. The cluster of Caenorhabditis is fairly diverged, but in the other species the colinear organization is strongly conserved. None of the Lineus Hox genes identified belong to the same ortholog group, which implies that Lineus also contains a single Hox cluster. This is supported by our preliminary analysis of the Lineus genome by pulse-field electrophoresis and Southern blotting (Fig. 5). If genomic DNA is cut with NotI and probed successively with LsHox3 and LsHox7, identical patterns of hybridization are observed with a single band in the range of 200–300 kb being radioactively labeled. A similar result is obtained if the DNA is partially digested with SacII. To test whether the pattern of hybridization is specific for each probe, we chose PstI digestion as an internal control, because only the LsHox3 probe contains a PstI site. As expected the LsHox3 probe yields two bands of hybridization when the DNA is cut with PstI, whereas a single band is observed when the DNA is probed with LsHox7 (Fig. 5). These data are consistent with the assumption of a single Hox cluster in Lineus, but the entire cluster has to be analyzed in detail or sequenced to confirm this hypothesis.

Figure 5.

Southern blot analysis of pulse-field gel. (A) Hybridization pattern obtained with the probe H3 (LsHox3 subclone). (B) Hybridization pattern obtained with the probe H7 (LsHox7 subclone). 1, uncut genomic DNA; 2, DNA digested with NotI; 3, DNA digested with PstI; 4, DNA partially digested with SacII. The same blot was hybridized successively with both probes. The blot was first hybridized with the probe H3. The autoradiogram was obtained after 12 hr of exposure. The blot was then dehybridized and hybridized again with the probe H7. The autoradiogram was obtained after 30 hr of exposure. The second hybridization has given signals of lower intensity and a higher background compared with the first hybridization. Both probes give the same hybridization pattern except in the case of DNA digested with PstI. This enzyme was used as a control because we expected, in this case, a different hybridization pattern with the two probes. The results confirm first, that no signal was left after dehybridization and second, that the hybridization patterns obtained for each probe are specific.

Recently it has been proposed that a common ancestor of protostomes and deuterostomes might have had a rather sophisticated body morphology elaborated under Hox and Pax gene control (42). To understand the evolution of animal architecture we have to understand the evolution of Hox genes and their role in the specification of body plans. Because the body plans of most phyla were already established at the Cambrian “explosion,” we have to study the more primitive phyla like cnidarians and “lower” triploblasts. Several hypotheses for Hox gene evolution have been put forward (12–14, 39–44). It has been proposed that even before cnidarian divergence there was at least one anterior-class (labial type), one middle-class (Antp type), and one posterior-class (Abdominal-B type) gene that arose from the duplication of a unique ancestral gene. So far, anterior- and middle-class Hox genes have been found in most protostomes studied. However, posterior class genes have only been identified in some arthropods (43, 44) and in Caenorhabditis (45). Thus, the existence of an ancestral representative of the posterior class is more questionable. However, this could be due to technical reasons, because the standard primers that are used do not detect posterior-class genes easily. In Lineus we have identified an Abdominal-B type gene, which shows significant sequence conservation compared with both arthropod and chordate ortholog group 9 genes. This gene was not detected by PCR methodology but only when the library was screened by hybridization at reduced stringency, indicating that it could easily have been missed. The identification of Abdominal-B type genes in Lineus, Caenorhabditis, Drosophila, and Amphioxus indicate that the posterior-class genes arose early in metazoan evolution. This is consistent with our hypothesis that Hox genes arose by repeated unequal crossing-over from a Ur-Hox gene, which first duplicated to form the terminal anterior- and posterior-class genes (12). The middle-class genes then would have been added progressively in between these two most ancient genes, which have most extensively diverged from the Ur-Hox gene. This hypothesis predicts that primitive Hox clusters should lack middle-class genes but contain the anterior (labial type) and posterior (Abdominal-B type) genes. Our findings in Lineus so far are consistent with this hypothesis. The Lineus cluster contains at least six genes (LsHox 1, 3, 4, 6, 7, and 9), including a representative of the anterior class (LsHox1) and the posterior class (LsHox9). We also have proposed that empty spiracles (and orthodenticle) and caudal, the two genes that share the YPWM motif with the Hox genes and are expressed at the extreme anterior and posterior ends of the embryo, respectively, might originally have been located at the ends of the cluster (12). Whether LsCdx, the caudal homolog, still forms part of the cluster (distal to LsHox9) remains to be determined. It is difficult to prove that a certain gene is absent from the cluster, but the available evidence would suggest that LsHox5 may be missing from the cluster. LsHox5 has not been identified in the screen, and a “chromosome walk” between LsHox4 and LsHox6 spanning 64 kb has not given any indication of a homeobox gene in between. However, all the Hox genes have to be mapped, and eventually the entire cluster will have to be sequenced to answer this point definitively. The possibility of a gene translocation also has to be considered. In fact, one additional Hox gene belonging to the Ubd-A (Ubx/abdA) group that is characteristic for the protostomes has been identified by using a different set of primers (A. Adoutte and G. Balavoine, personal communication). This gene would tentatively be assigned to LsHox8.

Because the branch point between nemerteans and chordates corresponds to the point of divergence between protostomes and deuterostomes, we propose that the last common ancestor of protostomes and deuterostomes, the urbilateria, had a Hox cluster with at least six or seven genes. The analysis of this cluster should provide further insights into the evolution of animal body plans. Even though Lineus is not easily amenable to genetic analysis, the extensive regeneration capacity and the development of grafting procedures (46) offer alternative possibilities for the study of Hox gene function in this animal.

ABBREVIATION

A-P

anterior–posterior