Parasegmental organization of the spider embryo implies that the parasegment is an evolutionary conserved entity in arthropod embryogenesis

The arthropod body consists of metameric units that become manifest as the segments at the germband stage (Anderson, 1973; Scholtz, 1997). The molecular mechanisms that underlie the segmentation process have been best studied in the insect Drosophila, where segmentation genes act in a hierarchic cascade; as a result, the metameric embryo is formed. A remarkable feature of Drosophila segmentation is that the fundamental developmental units are not the segments, but the parasegments that are defined by functional compartment boundaries (Martinez-Arias and Lawrence, 1985; Lawrence, 1988; Patel, 1994). The crustacean body is also initially built from units that resemble the parasegmental modules in insects (Patel et al., 1989a; Patel et al., 1989b; Patel, 1994; Dohle and Scholtz, 1988; Scholtz, 1997).

The subdivision of the anteroposterior body axis in the insect Drosophila results from the successive action of the maternal, gap, pair rule and segment polarity genes (Ingham, 1988; St Johnston and Nüsslein-Volhard, 1992; Pankratz and Jäckle, 1993). The pair rule genes delimit the parasegments, the initial metameric units in Drosophila, and define the domains that will express the segment polarity genes, such as engrailed and wingless (Lawrence et al., 1987; DiNardo and O’Farrell, 1987; Ingham, 1988; DiNardo et al., 1988; Baker, 1988). The engrailed gene encodes a homeobox-containing protein that is involved in establishing and maintaining the parasegmental boundaries in the Drosophila embryo. The anterior domain of the parasegment that expresses engrailed corresponds to the future posterior part of the segment in Drosophila as well as in other insects (Rogers and Kaufman, 1996; Schmidt-Ott et al., 1994; Patel et al., 1989a; Patel, 1994). Drosophila embryos in which engrailed is expressed uniformly are unsegmented (Lawrence et al., 1996). In addition, embryos that lack both wingless and engrailed function are unsegmented. The alternation of cells that express engrailed and non-expressing cells is essential for segmentation, and determines how these cells respond to morphogens (Lawrence et al., 1996).

In malacostracan crustaceans, Engrailed is expressed in the newly forming segments in the most anterior row of four rows of cells that form a genealogical unit (Patel et al., 1989a; Patel, 1994; Scholtz et al., 1994; Scholtz, 1995; Scholtz and Dohle, 1996). The row of Engrailed-expressing cells eventually ends up in the posterior region of each segment. The anterior part of the segment is formed from the posterior cells of the more anterior genealogical unit, which do not express Engrailed. These genealogical units correspond to units like the insect parasegments (Dohle and Scholtz, 1988; Patel, 1994).

The origin of segmentation in other arthropod groups like the chelicerates, a basal arthropod taxon, is still obscure. The chelicerates include the spiders, mites, scorpions and horseshoe crabs. Previous work suggests a role for the orthologs of the Drosophila pair rule genes hairy, even-skipped and runt in spider segmentation (Damen et al., 2000). These spider pair rule gene orthologs are expressed in a dynamic way in a domain at the posterior end of the embryo, from which stripes form. However, the exact mechanism that underlies chelicerate segmentation is still unclear. As the chelicerates form a basal arthropod group, characters in common between chelicerates and other arthropod taxa can be assumed as ancestral arthropod traits. The analysis of the segmentation process in chelicerates, therefore, may provide us with information on the basic embryonic molecular architecture of arthropods.

To obtain more insights into the evolution of developmental mechanisms that underlie the segmentation process in the arthropods, segment-polarity genes were studied in the spider Cupiennius salei (Chelicerata). Although there are differences, the expression of the spider engrailed genes, the wingless/Wnt genes and the cubitus interruptus gene imply that parasegmental boundaries are highly conserved within the arthropod clade.

Embryos of the Central American wandering spider Cupiennius salei Keys. (Chelicerata, Aranida, Ctenidae) were used (Damen et al., 1998; Damen and Tautz, 1998). Spiders were obtained from a colony bred by Ernst-August Seyfarth in Frankfurt am Main (Germany) or from our newly established colony in Cologne.

Fragments for spider genes were obtained by RT-PCR as described before (Damen et al., 2000). The oligo nucleotide primers used in the initial PCR for engrailed were en fw1 (TGGCCMGCMTGGGTNTWYTGYAC) and en bw-4 (TTRTAMARNCCYTSNGCCAT). In a nested PCR, the primers en fw-2 (GAMGAMAARMGNCCNMGNAC) and en bw-3 (RTTYTGRAACCADATYTTDATYTG) were used. For wingless, the primers wg-fw-1 (ATHGARWSNTGYACNTGYGAYTA) and wg-bw (ACYTWRCARCACCANTGRAANGTRCA) were used in the initial PCR, and wg-fw-2 (TGGGARTGGGGNGGNTGYWSNGA) and wg-bw were used in a nested PCR. For cubitus interruptus (ci) the oligonucleotide primers ci-fw (GARCANAAYTGYCAYTGG) and ci-bw-1 (CCRTGNACNGTYTTNACRTG) were used in the initial PCR, and ci-fw and ci-bw-2 (GGRTCNGTRTANCKYTTNG) in a nested PCR. The resulting PCR products were cloned and sequenced.

The obtained en PCR fragment was used to screen the embryonic C. salei cDNA library (Damen et al., 1998). One full-length clone (Cs-en-1) and three 5′ and/or 3′ truncated clones were isolated. Another engrailed cDNA clone, Cs-en-2, was recovered by screening the embryonic C. salei cDNA library under low stringency conditions with a probe for the homeodomain of Cs-abd-A (from position 410-615) (Damen et al., 1998). After an overnight hybridization at 52°C, the filters were washed twice with 2*SSC/0.1% SDS at 52°C for 15 minutes each. Several homeobox-containing genes were obtained (Damen et al., 1998) (W. G. M. D., unpublished), among them three cDNAs for Cs-en-2. The longest Cs-en-2 cDNA clone was sequenced.

The complete coding region of Cs-wg and Cs-Wnt5-1 were obtained by RACE-PCR (Marathon cDNA amplification kit, CLONTECH).

The sequences for the different genes were determined from both strands on an ABI-377XL automated sequencer (Applied Biosystems), using Big Dye dye-terminators (Perkin Elmer). The nucleotide data are available under Accession Numbers AJ007437 (Cs-en-1), AJ315944 (Cs-en-2), AJ315945 (Cs-wg), AJ315946 (Cs-Wnt5-1) and AJ315947 (Cs-ci)

In a test for wg genes in the spider, the following primers were used: initial PCR, wg-fwn1 (CAYAAYAAYGARGCNGG) and wg-bwn (CATNARRTCRCANCCRTC); nested PCR, wg-fwn2 (GARTGYAARTGYCAYGG) and wg-bwn.

Sequences were aligned using ClustalX (Thompson et al., 1994) and the BLOSUM matrix, a gap penalty of 20 and a gap extension of 0.2. Phylogenetic analysis was carried out using PUZZLE (Strimmer and von Haeseler, 1996) as implemented in PAUP 4.0b6 (Swofford, 2001).

Whole-mount in situ hybridization was performed essentially as described previously for Drosophila (Tautz and Pfeifle, 1989; Klingler and Gergen, 1993) with the modifications for spider embryos (Damen and Tautz, 1998; Damen and Tautz, 1999).

cDNAs for two different engrailed genes were recovered from Cupiennius salei. Cs-en-1 was found by RT-PCR and subsequent screening of the embryonic cDNA library (Damen et al., 1998), Cs-en-2 by low stringency screening of the same library with a Cs-abd-A homeodomain probe.

The 3416 bp Cs-en-1 cDNA contains an open reading frame (ORF) of 732 bp (position 133-864) and 2552 bp of 3′ UTR sequence with a polyadenylation signal and a short poly-A tail at its 3′ terminus. The deduced Cs-EN-1 protein is 244 amino acids long and includes an Engrailed-type homeodomain that is 67-82% identical to other Engrailed homeodomains (Fig. 1A). In addition, the Engrailed-specific-domains EH1-EH5 (Joyner and Hanks, 1991; Duboule, 1994) are recognized in the Cs-EN-1 sequence (Fig. 1B). A remarkable feature of Cs-EN-1 is the linker of 23 amino acid between EH2 and EH3, which is not found in other Engrailed sequences, where EH2 and EH3 are immediately adjacent to each other, except for a number of arthropod Engrailed/Invected proteins, which contain a two amino acids insertion (always Arg-Ser), and the amphioxus Engrailed protein, which contains a four amino acid insertion between EH2 and EH3 (Fig. 1B). The resemblance of the Cs-EN-1 homeodomain to the different Engrailed homeodomains and the presence of the Engrailed-specific-domains unambiguously show that Cs-en-1 is a spider engrailed ortholog.

The 1342 bp Cs-en-2 cDNA contains an ORF from position 1-470. The deduced and likely incomplete 156 amino acid protein contains an Engrailed-type homeodomain. However, this homeodomain is only 52-60% identical to other Engrailed homeodomains. The Engrailed-specific domains are also recognized in the Cs-EN-2 sequence (Fig. 1B). A remarkable point is the derived sequence of the highly conserved EH2 domain in both Cs-EN-1 and Cs-EN-2 (Fig. 1); nevertheless, the Cs-EN-2 homeodomain shows most similarities to Engrailed homeodomains in a BLAST search (Altschul et al., 1997). This similarity to other Engrailed homeodomains, the presence of the Engrailed-specific domains and many aspects of its expression (see later) suggest that Cs-en-2 is also an engrailed ortholog in the spider.

The spider ortholog of the wingless/Wnt1 gene was recovered by RT PCR and subsequent RACE PCR. The 3707 bp Cs-wg sequence contains an 1122 nucleotide ORF (position 16-1137). The likely full-length deduced protein encodes a 374 amino acid protein that clearly represents an ortholog of the Wingless/WNT1 class of WNT proteins, as becomes evident in a phylogenetic analysis (Fig. 2).

In addition to Cs-wg, four other members of the Wnt gene family were found in the spider: Cs-Wnt5-1 and Cs-Wnt5-2, two orthologs of the vertebrate Wnt5 gene (DWnt3/5 in Drosophila); and Cs-Wnt7-1 and Cs-Wnt7-2, two orthologs of the vertebrate Wnt7 gene (DWnt2 in Drosophila). The Cs-Wnt-5-1 gene appears to have an interesting expression pattern with respect to segmentation. Therefore, a 2148 nucleotide sequence for Cs-Wnt5-1 was recovered by RACE-PCR containing an 1143 nucleotide ORF (position 265-1407). The deduced Cs-WNT5-1 protein is a 381 amino acid protein. Cs-WNT5-1 clusters with vertebrate WNT5 and Drosophila DWNT3/5 in a phylogenetic analysis (Fig. 2).

A spider ortholog of the Drosophila cubitus interruptus (ci) gene was isolated by RT-PCR. The PCR product representing the spider Cs-ci gene is 391 base pairs long. The 130 amino acid Cs-CI protein fragment deduced from this sequence corresponds to amino acids 446-579 of the Drosophila CI protein. The Cs-CI fragment is 86% identical to the corresponding domain in the Drosophila CI protein, and up to 84% identical to the corresponding domain in vertebrate GLI proteins.

Some aspects of the expression of Cs-en-1 in spider embryos have been described previously (Damen et al., 1998), where its expression was used as a segmental marker in the spider embryo. The current paper describes all aspects of Cs-en-1 expression, and, in addition, the embryonic expression of the Cs-en-2 gene, as well as that of wingless, Wnt5-1 and cubitus interruptus.

To allow a better understanding of the expression patterns in the spider, a short introduction is given to some morphological features. Chelicerates have two tagmata: a prosoma and an opisthosoma. The prosoma is the cephalothorax and bears six pairs of appendages in the spider: the cheliceres, the pedipalps and four pair of walking legs. The opisthosoma is the ‘abdomen’ of chelicerates. In spiders, opisthosomal limb buds (appendage anlagen) form on the second to fifth opisthosomal segment. These opisthosomal limb buds will form the respiratory organs (book lungs, first pair of buds) and the spinnerets (third and fourth pair) (Foelix, 1996).

In very early germ band stage embryos, Cs-en-1 is expressed in five clear stripes, representing the pedipalp and the four walking leg segments; additionally, a very weak stripe is seen where the chelicere segment is forming (Fig. 3A). This chelicere segment is the anterior-most appendage-bearing segment and forms a little later than the other prosomal segments (Seitz, 1966). Somewhat later, this cheliceral Cs-en-1 stripe is stained as strongly as the other prosomal stripes; in addition, the Cs-en-1 stripes widen (Fig. 3B-F). Newly formed Cs-en-1 stripes become successively visible posterior to the last prosomal segment, demarcating the first segments of the opisthosoma (Fig. 3D,F). Initially, these opisthosomal stripes seem to be somewhat narrower compared with the stripes in the prosomal segments. However, these stripes soon widen. Additional Cs-en-1 stripes form successively at the posterior end of the embryo. At the stage when two opisthosomal stripes are visible, two spots of Cs-en-1 expression become visible in the head region, anterior to the cheliceral Cs-en-1 stripe (Fig. 3C). Later, these spots transform into small stripes. These spots probably demarcate the ocular segment that corresponds to the ocular (or pre-antennal) segment in insects and crustaceans, as recognized previously (Damen et al., 1998).

As soon as appendages form (Fig. 3E,F), Cs-en-1 is primarily expressed in the ventral part of the embryo, which becomes even more prominent at later stages. Additionally, there is Cs-en-1 expression in the posterior part of the appendages themselves. No Cs-en-1 expression is visible dorsal to the prosomal appendages, except to the cheliceres. Only weak Cs-en-1 expression is visible at the dorsal region of the opisthosomal segments (Fig. 3G,H,J).

At the so-called inversion stage, which results in dorsal closure, there are up to twelve Cs-en-1 stripes detectable in the opisthosoma. The most posterior opistosomal segments appear especially to be very small; the engrailed stripe in the twelfth segment is only visible after DAPI counterstaining (Fig. 3K). Somewhat later, a pro-larval stage has been reached and the ring-like expression of Cs-en-1 becomes obvious at the posterior end (Fig. 3L). This ring-like structure resembles the ring structure in a number of insects and may correspond to the proctodeum expression (Schmidt-Ott et al., 1994). From the inversion stage onwards, weak Cs-en-1 expression is visible anterior to the labrum (Fig. 3I).

The expression of Cs-en-2, the second engrailed gene in Cupiennius, deviates from that of Cs-en-1. The expression of Cs-en-2 becomes apparent in a double stripe fashion somewhat later than Cs-en-1 expression (Fig. 4A,B). The opisthosomal Cs-en-2 stripes seem to split off from a larger domain of expression at the very posterior end of the embryo (Fig. 4C,E). These newly formed Cs-en-2 stripes are also doublets; however, the cells between this doublet stripe express low levels of engrailed as becomes apparent after elongated staining (not shown). As soon as the limb buds appear as a landmark (Fig. 4D,E), it becomes evident that the anterior stripe of each double stripe marks the same anterior boundary as does Cs-en-1. However, Cs-en-2 is not expressed in the posteriormost part of the appendage, whereas Cs-en-1 is expressed in the complete posterior portion of the appendages (Fig. 4H). The posterior stripe of each doublet is located just posterior to the appendages, obeying a similar posterior border as Cs-en-1, although it is not possible to determine whether these posterior borders are identical, owing to the lack of a positional marker here.

Similar to Cs-en-1, Cs-en-2 is expressed in a ring-like structure at the posterior end of the embryo (Fig. 4I), as well as anteriorly to the labrum (not shown). In contrast to Cs-en-1, Cs-en-2 is expressed in the prospective stomodeum in early germ bands (Fig. 4F); at later stages, this form a ring in the foregut.

Both engrailed genes are expressed in the limb buds and the appendages that form from them. The appendages on the prosomal segments (Fig. 5B, Fig. 3E,F, Fig. 4D,F), as well as the opisthosomal limb buds (second to fifth opisthosomal segment), straddle the anterior boundary of Cs-en-1 and Cs-en-2 expression (Fig. 3G, Fig. 4G), suggesting that this boundary is an important developmental boundary.

Another remarkable aspect of the engrailed expression in the spider is the observation that the anterior border of the engrailed stripes sharpens earlier compared with the posterior border of each stripe (Fig. 5A). This may be the result of the action of wg/Wnt genes. In Drosophila, wg-expressing cells reside directly anteriorly to en-expressing cells, and a sharp parasegment boundary is formed between wg- and en-expressing cells, which are mutually exclusive. The wg and en genes function to maintain the polarity of the segments in insects (Martinez-Arias et al., 1988; Van den Heuvel et al., 1989; Nagy and Carroll, 1994; Oppenheimer et al., 1999).

The expression of en is activated by the action of the pair rule genes in Drosophila. In a second phase, en expression becomes autocatalytic, but is also influenced by wg. Later in Drosophila development, en expression becomes independent of wg (Heemskerk et al., 1991). The role of wg seems to be conserved in insects and crustaceans (Nagy and Carroll, 1994; Nulsen and Nagy, 1999; Oppenheimer et al., 1999). To test whether the signaling between en- and wg-expressing cells is present in the spider, members of the wg/Wnt gene family from the spider have been analyzed.

The Cs-wg gene is expressed in a segmentally iterated pattern in the spider embryo (Fig. 6). Expression is first detected after Cs-en-1 and Cs-en-2 expression can be detected. In the prosomal segments, Cs-wg is initially expressed only in a stripe in the anteroventral region of the appendages (Fig. 6A,E). The posterior expression border lies in the middle of the appendages, just anterior to the anterior border of engrailed expression. Unfortunately, it is not yet possible to double stain for these genes in the spider embryo to verify that the expression domains for en and wg in the spider are touching each other, as is the case in Drosophila. Nevertheless, the position of the anterior en and the posterior wg expression border just in the middle of the appendages strongly suggests that these expression domains are adjacent to each other.

Later, a spot of Cs-wg becomes visible dorsal to the prosomal appendages (Fig. 6D). In the opisthosoma, Cs-wg is expressed in small stripes at the dorsal side of the newly formed segments (Fig. 6B). These stripes expand later (Fig. 6C,F,H) and are just dorsal to the opisthosomal limb buds, but do not expand completely to the dorsal side of the germ band (Fig. 6G). In addition, there is a small spot of Cs-wg expression just ventral to the opisthosomal limb buds, but not in the neuroectoderm more ventrally. The limb buds are good landmarks for demonstrating indirectly that the Cs-wg expression is just anterior to engrailed expression (Fig. 3G,H). Remarkably, the posterior margin of Cs-wg expression, just anterior to the sharp engrailed margin, forms a sharp border (Fig. 6C,F,G).

There is also a spot of Cs-wg expression posteriorly, in the abdominal limb buds on O2, O4 and O5, but not on O3 (Fig. 6F,G, arrowhead). This spot is in the Cs-en-1 domain that covers the complete posterior part of the limb buds.

Although there is expression of Cs-en-1 in the ventral neuroectoderm, there is no adjacent expression of Cs-wg here. To test whether there is a second wg gene in the spider that might function in this region of the embryo, RT-PCR was performed. Degenerated primers were used that lie in other domains than the ones used in the cloning of the Wnt genes (see Materials and Methods section) on RNA from an early germband stage spider (limb buds just forming). At this stage, segmentation takes place, and one would expect the gene involved in the segmentation process to be expressed. No additional genes were found in this PCR screen. Forty-three clones were sequenced: seven corresponded to Cs-wg-1, nine to Cs-Wnt5-1, six to Cs-Wnt5-2, eight to Cs-Wnt7-1 and 13 to Cs-Wnt7-2. Although this does not form indisputable evidence, it is not very likely that there is a second wg/Wnt1 gene in the spider. This is supported by the expression of the Cs-Wnt5-1 gene, which might act in the ‘missing’ domain (see below).

In addition to the segmental expression, Cs-wg is expressed in two spots in the head (Fig. 6A,E), in a small stripe anterior in the labrum (Fig. 6E) and at the posterior end of the embryo (Fig. 6B,H). A comparable posterior domain is found in embryos of Drosophila, Tribolium (beetle) and Triops (branchiopod, crustacean) (Baker, 1988; Nagy and Carroll, 1994; Nulsen and Nagy, 1999). It has been proposed that the posterior wg-expressing cells could act as a source for a morphogen necessary for the function of the growth zone (Nulsen and Nagy, 1999).

Surprisingly, the Cs-Wnt5-1 gene shows a segmental expression in those regions of the embryo where wg expression is expected but where Cs-wg is not expressed (Fig. 7). The Drosophila ortholog DWnt3/5 does not have a segmental function (Fradkin et al., 1995; Klingensmith and Nusse, 1994).

Cs-Wnt5-1 is expressed ventrally to the appendages and the opisthosomal limb buds, and also in an identical position in the segments that do not bear appendages (Fig. 7A-E). Although its posterior expression border is not as sharp as that of Cs-wg, the posterior border is just anterior to the position of the engrailed expression domains, again using the appendages as a landmark (Fig. 7E,F). The Cs-Wnt5-1 gene, therefore, may act in this ventral region as a segment polarity gene. An in situ hybridization with both the Cs-wg and Cs-Wnt5-1 probe shows that the two genes together cover the whole width of the germ band (Fig. 7H), just interrupted by the appendage anlagen, suggesting that both genes act similarly but in a different domain along the dorsoventral axis. Furthermore, Cs-Wnt5-1 is expressed in segmentally dorsal spots in the opisthosoma (Fig. 7F,G), in rings in the appendages (Fig. 7F), in four spots in the head, in the labrum (Fig. 7A) and weakly at the very posterior end (Fig. 7C-E).

Additional evidence for the conservation of the Engrailed-Wingless/Wnt pathway comes from expression of the cubitus interruptus (ci) ortholog in the spider. Ci is a transcriptional activator for wg expression in Drosophila, and is expressed in the cells that do not express engrailed (Eaton and Kornberg, 1990; Motzny and Holmgren, 1995; Aza-Blanc and Kornberg, 1999). The spider Cs-ci gene is also expressed in the segmental regions that do not express engrailed, in a similar way to its ortholog in the fly (Fig. 8).

These data show that some of the major players that establish the parasegment boundary in Drosophila are also present in the spider, and that the segment-polarity network that establishes and maintains the parasegment boundary is likely to be present in the spider.

Expression of segmentation genes point to a parasegmental organization. The presumptive parasegment borders are defined by grooves and are morphologically visible in the spider embryo, as in the fly embryo (Lawrence, 1992). Metamerization becomes morphologically visible in the spider embryo as soon as grooves form, as visualized by DAPI staining in Fig. 5E. As the anterior border of engrailed expression (Fig. 5C,D) and the posterior border of Cs-wg and Cs-Wnt5-1 expression (Fig. 6B, Fig. 7D,E) are confined to the edge of these grooves, this suggests that the grooves define a parasegmental, rather than a segmental, subdivision.

Genetic and molecular studies have shown that parasegments are fundamental units in the design of the Drosophila embryo (Martinez-Arias and Lawrence, 1985; Lawrence et al., 1987; Lawrence, 1988; Lawrence, 1992). In addition, molecular comparisons demonstrate that parasegments are almost certainly the fundamental units of development not only in insects, but also in crustaceans (Dohle and Scholtz, 1988; Patel, 1994). Several pieces of evidence demonstrate that the spider embryo is presumably also built from parasegmental units. The parasegment, thus, is probably an entity that is evolutionarily conserved in arthropods.

In the spider, the appendages straddle the anterior border of the engrailed expression domain, as in insects and crustaceans, where they match exactly to the borders of parasegmental boundaries (Patel et al., 1989a; Martinez-Arias, 1993; Scholtz, 1995; Dohle and Scholtz, 1988). Boundaries play an important role as organizers. The appendages start forming at the intersection of the anteroposterior and dorsoventral boundaries, as has been demonstrated by producing ectopic boundaries (Cohen, 1993; Cohen et al., 1993; Tabata et al., 1995; Serrano and O’Farrell, 1997; Niwa et al., 2000). At this intersection of boundaries, both Wg and DPP are produced, and their synergistic activity determines an organizer for appendage formation. It is not yet known how the dorsoventral axis is determined in the spider. Nevertheless, the formation of appendages on the border defined by engrailed indicates that this border specifies a functional compartment boundary for appendage formation in the spider.

An additional piece of evidence comes from the observation that the anterior border of the engrailed stripes sharpens earlier than the posterior border of each stripe, as in insects and crustaceans (Patel, 1994). The posterior margin of Cs-wg expression that is adjacent to the sharp engrailed expression border is a sharp border as well. This implies that in chelicerates the parasegments are also the first metameric units to be resolved.

Another important argument for the parasegmental organization of the insect embryo is that key developmental genes are expressed in such domains (Struhl, 1984; Martinez-Arias and Lawrence, 1985; Lawrence, 1988) (summarized in Fig. 9). The chelicerate posterior Hox genes (Antennapedia, Ultrabithorax-2, abdominal A and Abdominal-B) obey anterior expression borders (Damen et al., 1998; Telford and Thomas, 1998; Damen and Tautz, 1999) that correspond to boundaries defined by engrailed, as in insects (Struhl, 1984; Kaufman et al., 1990), and imply the existence of a functional parasegmental organization (Fig. 9).

By contrast, the anterior Hox genes (labial, proboscipedia, Hox3, Deformed and Sex comb reduced) are expressed in a segmental register rather than a parasegmental one in both chelicerates and insects (Kaufman et al., 1990; Damen et al., 1998; Telford and Thomas, 1998; Damen and Tautz, 1999; Abzhanov et al., 1999) (M. Schoppmeier and W. G. M. D., unpublished data). Remarkably, the anterior expression borders of most of these anterior Hox genes lie outside the region that, in Drosophila, is patterned as a result of the action of the pair rule genes and the engrailed stripes that directly depend on these pair rule genes (Fig. 9). The patterning of this anterior head region is probably controlled in a different way (Gallitano-Mendel and Finkelstein, 1997). Although pair rule gene orthologs might be involved in spider segmentation (Damen et al., 2000), it is not known yet whether these genes act in comparable regions of the embryo, as in Drosophila.

There is an ongoing discussion of whether segmentation in different phyla has a common origin (Davis and Patel, 1999). The presumably conserved segment-polarity network and the organization into parasegments can be seen as an ancestral character for arthropods. In the closely related onychophorans, engrailed expression points to a comparable organization (Wedeen et al., 1997). However, segment polarity gene orthologs are apparently not involved in body segmentation in other segmented phyla. In annelids, engrailed is expressed in segmentally iterated spots in the CNS and in mesodermal cells, but is probably not involved in body segmentation as in arthropods (Wedeen and Weisblat, 1991; Lans et al., 1993; Seaver and Shankland, 2001; Seaver et al., 2001). The establishment of segment polarity in leeches is independent of cell interactions along the anteroposterior axis; this is in contrast to the situation in arthropods, where anterior and posterior fates of the segments are specified by intercellular signaling between wg- and en-expressing cells. (Seaver and Shankland, 2001). Furthermore, there are no indications that the annelid embryo is constructed from units like the parasegment. In the leech, progeny of particular teloblasts overlap with respect to segmental boundaries and do not form genealogical units like in crustaceans (Weisblat and Shankland, 1985; Irvine and Martindale, 1996). Some key aspects of arthropod segmentation are thus not present in annelids. The segmentation of annelids and arthropods, therefore, seems to be brought about by different mechanisms. This is an important argument against a common origin of segmentation in annelids and arthropods.

In chordates it is also doubtful whether engrailed plays a role in somitogenesis. engrailed but not wingless is expressed in reiterated pattern in the somites of the cephalochordate amphioxus (Holland et al., 1997; Holland et al., 2000), which suggests that the segment polarity gene network as present in arthropods is not conserved. Furthermore, vertebrate engrailed orthologs do not play a role in somite formation or maintenance of the somite boundaries. This points to different mode of segmentation in vertebrates and arthropods, and does not support a common origin of segmentation. However additional evidence is required to prove this.

In several metazoan phyla there are representatives that contain duplicated engrailed genes, whereas others contain only one gene, pointing to independent duplications. Duplicated engrailed genes have been found in several insect groups, like the two engrailed paralogs engrailed and invected in Drosophila (Coleman et al., 1987; Hui et al., 1992; Peterson et al., 1998; Marie and Bacon, 2000), whereas in insects such as Tribolium and Schistocerca, only one engrailed gene has been detected (Patel et al., 1989a; Brown et al., 1994). Independent duplications of the engrailed gene also appear to have taken place in some crustacean lineages (Gibert et al., 1997; Gibert et al., 2000; Abzhanov and Kaufman, 2000). The same is known from other phyla, as in some molluscs (Wray et al., 1995) and chordates (Joyner and Martin, 1987; Joyner and Hanks, 1991; Holland and Williams, 1990; Holland et al., 1997).

In the spider, two engrailed genes have been found; however, phylogenetic analyses (not shown) do not allow conclusions on the origin of the duplication. Only one engrailed gene has been described for another chelicerate, the mite Archegozetes longisetosus (Telford and Thomas, 1998), which suggests that the duplication of engrailed genes in chelicerates is restricted to the spider lineage. However, the spider Cs-en-2 gene was not found in our PCR screen with redundant primers, probably owing to sequence derivation of the Cs-en-2 EH2 domain (see Fig. 1) to which the PCR-primers were directed. The PCR method was also used to find the mite engrailed ortholog (Telford and Thomas, 1998). Therefore, a second engrailed gene could be missed in the PCR screen for the mite, as was the case for Cs-en-2 of the spider. Nonetheless, a duplication of the engrailed gene took place somewhere in the chelicerate lineage. It remains to be elucidated whether this duplication took place before or after the spiders and mites diverged.

The two spider engrailed genes both seem to define the same boundary; nevertheless, the way they appear is very different and suggests different modes of regulation. Cs-en-1 is expressed in a comparable way to engrailed in insects and crustaceans. Its expression starts in the region where expression of the spider orthologs of the Drosophila pair rule genes hairy, even-skipped and runt diminishes (Damen et al., 2000). It is not yet possible to produce double labeling in the spider; nevertheless, this correlation suggests that the pair rule gene orthologs may act upstream of the Cs-en-1 gene in the spider, as is the case in insects where the engrailed expression domains are defined by the action of the pair rule genes (DiNardo and O’Farrell, 1987; DiNardo et al., 1988; Patel et al., 1994; Rohr et al., 1999).

However, both the expression of Cs-en-2 at the most posterior end of the embryo and the doublet stripes are atypical and unique for engrailed genes. The way the Cs-en-2 stripes form is not completely clear; they seem to originate from the broad posterior domain and than split to form the doublet (Fig. 4E). Nonetheless, the final anterior position of the anterior stripe of the doublet seems to be identical to the ones for Cs-en-1 and might also be maintained by interaction with Cs-wg/Cs-Wnt5-1-expressing cells.

The broad posterior domain of Cs-en-2 expression in the spider embryo is in a comparable domain to the spider pair rule orthologs (Damen et al., 2000), giving some indication that the Cs-en-2 gene might act as a more upstream segmentation gene. However, in contrast to the spider pair rule gene orthologs hairy, even-skipped and runt (Damen et al., 2000), the expression of Cs-en-2 is not dynamic in this posterior domain. However, Cs-en-2 expression is only detected after Cs-en-1 expression in the early germ band stages when the prosomal segments form. Thus, there might be a difference between the specification of the prosomal segments and the opisthosomal segments that are formed from the posterior growth zone. Further analysis of the Cs-en-2 gene is required to answer these questions.

During the course of development, the two spider engrailed genes predominantly act in different domains along the dorsoventral axis. At the onset of inversion, Cs-en-1 is less intensively expressed at the future dorsal side, whereas expression of Cs-en-2 is completely reduced at the future ventral side. By contrast, the duplicated insect engrailed genes are expressed in more or less redundant domains (Coleman et al., 1987; Peterson et al., 1998; Marie and Bacon, 2000), whereas the duplicated crustacean engrailed genes have different modes of expression (Gibert et al., 2000; Abzhanov and Kaufman, 2000). However, these differences are not as dramatic as the ones seen in the spider. The spider wg/Wnt class genes Cs-wg and Cs-Wnt5-1 are also differently expressed along the dorsoventral axis of the embryo and together they appear to cover the complete dorsoventral axis.

In Drosophila, cells along the dorsoventral axis acquire stable en expression at different times and no longer need wg function for en expression (Bejsovec and Martinez Arias, 1991). This transition of en regulation happens first at the dorsal side of the embryo and later also at the ventral side of the embryo, and is even reflected in dorsoventral differences in activity of the en promoter (DiNardo et al., 1988). The different modes of regulation of the engrailed gene in Drosophila along the dorsoventral axis of the embryo might be reflected in the differential expression along the dorsoventral axis of the spider engrailed genes, as well as the spider wg/Wnt5-1 genes.

The Cs-Wnt5-1 gene is probably involved in segmentation. The Cs-Wnt5-1 expression pattern suggests that the gene acts in the ventral region of the germ band as a segment-polarity gene in a domain where the Cs-wg gene is not expressed. In insects (Drosophila, Tribolium and the cricket Gryllus) and the crustacean Triops, the wg gene seems to cover the complete width of the germ band (Baker, 1988; Nagy and Carroll, 1994; Nulsen and Nagy, 1999; Niwa et al., 2000). The Drosophila ortholog of Cs-Wnt5-1, DWnt3/5, does not have a function in segmentation, whereas crustacean Wnt5 orthologs have not yet been analyzed (Fradkin et al., 1995; Klingensmith and Nusse, 1994). The spider Cs-wg gene and Cs-Wnt5-1 gene thus seem together to perform the function of the single wg gene in insects or crustaceans. It remains open to speculation whether the segmental role of Cs-Wnt5-1 in the spider is one that has been acquired in the chelicerate lineage and replaces the function of wg in this region, or one that has been lost in the lineage leading to the insects and crustaceans, and has been replaced there by wg.

Seitz (Seitz, 1966) recognized in his morphological description of the C. salei embryo, nine segments in the opisthosoma of the developing spider embryo. However, the 12 engrailed stripes as well as 12 segmentally iterated spots of both Cs-Pax6 and Cs-prd-1 (W. G. M. D., unpublished) in the opisthosoma of the Cupiennius embryo points to 12 opisthosomal segments. Twelve opisthosomal segments probably represents the ancestral state for spiders, and for chelicerates in general (Foelix, 1996; Westheide and Rieger, 1996). Mesothelae, the phylogenetically oldest spiders, still contain a segmented opisthosoma that consists of 12 metameres (Foelix, 1996). This is in contrast to more advanced spiders, like Cupiennius, where the segmentation of the opisthosoma is obvious only in embryos. These data thus show that, although morphologically hardly detectable, the opisthosoma of Cupiennius consists of 12 segments, which represents the ancestral state for spiders and chelicerates.

I thank Diethard Tautz for continued support, Ernst-August Seyfarth (Frankfurt am Main) for providing us with fertilized adult spiders in an initial phase of the research, Diethard Tautz and Hilary Dove for critical reading of the manuscript, Gabi Büttner for assistance with many of the situ hybridization experiments, and Barbara Wigand for performing the RACE-PCRs of Cs-Wnt5-1. This work was supported by a Marie Curie Fellowship of the European Commission (Brussels) and a DFG grant (Da526/1-1).