We isolated fragments of proboscipedia (Cs-pb), two copies of Sex comb reduced (Cs-Scr-1 and Cs-Scr-2), and a second copy of Deformed (Cs-Dfd-2) from the spider Cupiennius salei. Alignment with chelicerate and other arthropod sequences unambiguously show that these are Cupiennius orthologs of these Hox genes (Fig. 1). pb and Scr class Hox genes have not been recovered in Cupiennius before, but some data are available from other chelicerates: the mite Archegozetes longisetosus, the common house spider Achaearanea tepidariorum and the seaspider Endeis spinosa 141921. However, with the isolation of pb and Scr from Cupiennius, this spider is the first chelicerate species for which orthologs of all ten arthropod Hox gene classes are described.

More importantly, it becomes clear that at least three Hox genes in Cupiennius are present as duplicate copies. There are two Dfd orthologs [15, this paper], two Scr orthologs (this paper), and two Ubx orthologs 15. Despite the similarities in the amino acid sequence and especially within the homeodomain, the two copies are significantly different from each other on the DNA sequence level (not shown) and thus are different genes and not different alleles. At the moment it is unclear whether there are additional copies of other Cupiennius Hox genes. Abzhanov et al 14 also described two Dfd genes for another spider (Achaearanea tepidariorum), however, for one of them they only obtained a small PCR fragment encoding 27 amino acids within the highly conserved homeodomain. We recovered additional sequence information of this Achaearanea Dfd gene (At-Dfd-1) via RACE-PCR (Fig 1). The Cs-Dfd-1 15 and At-Dfd-1 sequences are more similar to each other than to the newly identified Cs-Dfd-2 sequence or to the At-Dfd-2 14 sequence. Also Cs-Dfd-2 and At-Dfd-2 are more similar to each other than to Cs-Dfd-1 or At-Dfd-1. This is most obvious from the sequence between hexapeptide and homeodomain. Based on the sequences (Fig. 1) we propose that Cs-Dfd-1 and At-Dfd-1 are gene orthologs and that Cs-Dfd-2 and At-Dfd-2 are gene orthologs. Also the expression patterns of At-Dfd-1 14 and Cs-Dfd-1 15 in the legs are remarkably similar (see also below). Therefore, the duplication of Dfd presumably was already present in the last common ancestor of these two spiders.

There is another arthropod, the geophilomorph centipede Pachymerium ferrugineum (Myriapoda), that contains two copies of the Dfd gene 16. However, these two centipede Dfd genes are more similar to each other than to any of the spider genes (Fig. 1). In addition, no Hox gene duplications have been described for other myriapods, e.g. the centipede Lithobius atkinsoni 22 and the millipede Glomeris marginata 23. This suggests that the two Dfd genes in Pachymerium are the result of an independent duplication event in the geophilomorph centipedes.

The expression of the Hox genes was studied via in situ hybridizations. Cs-pb is expressed in the pedipalpal segment and the four walking leg segments (L1-L4) (Fig. 2). This is similar to the common house spider Achaearanea and the mite Archegozetes longisetosus 1420. Cs-pb thus is expressed in the same segments as Cs-lab and Cs-Hox3 1517 (see also summary in Figure 6). At the limb bud stage (Fig. 2A) the expression is most obvious in the appendages (pedipalps and walking legs) but there is also some expression in the ventral ectoderm. At the inversion stage (Fig. 2B) Cs-pb expression is also clearly visible in the ventral ectoderm. We never observed any expression in the cheliceral segment or the opisthosomal segments.

Expression of the Cs-Dfd-2 gene is limited to the four segments that bear the walking legs (L1-L4) (Fig. 3D–F). These are the same segments that express the Cs-Dfd-1 gene (Fig. 3A–C). However, there are differences in the intrasegmental domains of the expression of the two Cs-Dfd paralogs. Neither are homogenously expressed, but each gene is expressed in a distinct pattern within the leg segments. Most prominent is the very strong expression of Cs-Dfd-1 at the most distal tip of the legs. Although Cs-Dfd-2 also is expressed in the distal tip, this expression is not as prominent as the one of Cs-Dfd-1. Furthermore, while Cs-Dfd-1 is expressed in all four walking legs at the same intensity (Fig. 3C), expression of Cs-Dfd-2 is weaker in L3 and L4 compared to L1 and L2 (Fig. 3F). Another difference is the strength of expression in the ectoderm ventral to the legs: Cs-Dfd-2 is only weakly expressed, while Cs-Dfd-1 is strongly expressed here (compare Fig 3A and Fig 3D). The common house spider Achaearanea tepidariorum also contains two Dfd genes 14. Comparison of the expression pattern of the two Cupiennius Dfd genes with the two Achaearanea Dfd genes 14 shows that Cs-Dfd-1 and At-Dfd-1 show similarities in their expression patterns. Most typical is the strong expression at the distal tip of the leg, which is much less prominent for Cs-Dfd-2 and At-Dfd-2. This prominent expression in the tip of the leg of Cs-Dfd-1 is most obvious when the colour reaction of the in situ develops (not shown).

Cs-Scr-1 and Cs-Scr-2 also are expressed in similar but not identical patterns. Cs -Scr-1 is initially expressed in the second, third and fourth walking leg segment (L2-L4) (Fig 4A). In the appendages, expression first appears only in the walking legs of L3 and L4 (Fig. 4B) and only later, but weaker, in the walking legs of L2 (Fig 4C). Cs-Scr-2 is also initially expressed in L2-L4, but the expression is not as widespread as the Cs-Scr-1 expression is as it is restricted to some small spots in the ventral ectoderm (Fig. 4D). Later expression is seen in the legs of L3 and L4 (Fig. 4E) but we never observed expression of Cs-Scr-2 in the legs of L2 (Fig. 4F). As with the two Dfd genes the patterns within the legs also differ for the two Cupiennius Scr genes (compare Fig. 4C and Fig. 4F).

The expression patterns for Cs-Ubx-1 and Cs-Ubx-2 have been described before 15. The anterior border of Cs-Ubx-1 is slightly more anterior than that of Cs-Ubx-2 15. There are small intrasegmental expression differences between the two Ubx genes. Cs-Ubx-2 is more homogeneously expressed compared to Cs-Ubx-1 (Fig 5A,B).

The combined data from other arthropods [e.g. 81920], summarized in Hughes and Kaufman 12, imply that the Hox complex of the last common ancestor of all arthropods contained ten Hox genes. The present data of the Cs-pb and Cs-Scr genes combined with our previous work 15171819 show that at least one copy of each of the ten arthropod Hox genes is present in the spider Cupiennius (Fig. 6). At least three of the Hox genes are even present as two copies (see below). The expression data on Cs-pb and Cs-Scr make Cupiennius the first chelicerate for which expression data are known for all ten different arthropod Hox genes; previously the chelicerate data were an assemblage from three different species [see also Ref. 12].

Our data show that at least three Hox genes are present as two copies in Cupiennius [combined data from this paper and Ref. 15]. There are previous reports on duplication of Hox genes in chelicerates. Cartwright et al. 13 could identify one to four representatives per Hox gene class in the horseshoe crab Limulus polyphemus. However, there is no expression data for the Limulus Hox genes. In the spiders Achaearanea 14 and Cupiennius 15, previous one duplicated Hox gene each had been described. For mite and pycnogonids no duplicated Hox genes have been described [e.g. 2021].

In all three cases in Cupiennius (Dfd, Scr, and Ubx), the two paralogs are expressed in comparable but not identical domains. They are expressed in the same segments with differences in the intrasegmental patterns. This shows a striking similarity with what is seen for the duplicated Hox genes of the various paralogous groups in vertebrates that are also expressed in similar but not completely identical expression domains 24.

Gene duplications offer several possible outcomes 25. One option is that one of the copies gets silenced or lost again during evolution. A second option is that one copy retains the ancestral function, freeing the other copy to diverge and evolve new functions (neofunctionalization). A third possibility is that each of the two copies performs a different subset of the ancestral functions (subfunctionalization). The differences in the intrasegmental expression patterns of the two copies in Cupiennius suggest that each of the two copies performs different functions that must be the result of either a neofunctionalization or a subfuntionalization event. As there is no genomic information available yet, it remains unclear whether there are additional duplicated Hox genes in Cupiennius (see also next section).

An important question is why duplicated Hox genes are present in the spider and why they are retained? Are they remnants of a large duplication event that are maintained because of neofunctionalization or subfunctionalization events? Or is there another evolutionary advantage for the spider to have multiple copies of some Hox genes? Presently it is difficult to answer these questions. In chelicerates there seems to be a tendency towards having more Hox genes [this paper, 131415], this in contrast to insects where there is a reduction of true Hox genes as two Hox genes -Hox3 and ftz- lost their homeotic function and obtained new functions in the insect embryo, which is associated with a divergence of the sequence of the gene 262728.

Hox gene duplications have been proposed to be one of the genetic mechanisms behind the diversification of vertebrates [e.g. 29]. However it remains difficult to draw a direct link between Hox gene duplications and morphological evolution. Recent results from Lynch et al 30 suggest an important role for the action of positive Darwinian selection in the divergence of vertebrate Hox genes after cluster duplications. The locations in the homeodomain of the sites that are under positive selection suggest that they are involved in protein-protein interactions. This suggests that adaptive evolution actively contributed to Hox gene function 30. Indeed, in the Cs-Ubx-2 homeodomain there are two amino acid exchanges compared to the homeodomain of Cs-Ubx-1 or of Ubx of most other arthropods (Fig 1). Only in the honey-bee and the crustaceans Moina and Artemia there is one amino acid exchange in the homeodomain, in all three cases an A to S exchange at position 37 of the homeodomain (not shown). Also one of the two exchanges in Cs-Ubx-2 is an A to S on position 37. The sequence divergence in the homeodomain of Cs-Ubx-2 thus might be associated with a functional divergence. However the mechanism of the divergence is unknown, leaving open the role of Hox gene duplication in morphological evolution of chelicerates.

The most important question that comes up now is on the origin of these three duplicated Hox genes in Cupiennius. There are two options. First, they result from a duplication of the complete cluster. This would imply that either additional Hox genes are present as two copies that have not be found so far, or that one copy has been lost for the other Hox genes, as has happened to some of the Hox genes in the duplicated vertebrate clusters. Mammals for instance possess four Hox clusters, but most of the paralogs are not present as four copies as some of them have been lost in some of the clusters 11. All data for Cupiennius Hox genes were obtained via either PCR approaches or cDNA library screening 15. As there is no genome project for the spider yet, this means that it is presently unclear whether additional Hox genes are present as duplicated copies in the spider. The second possible explanation for the three duplicated Cupiennius Hox genes could be three independent tandem duplications of the individual genes. Additional analyses are required to identify the genomic organization of the spider Hox genes, and to find out whether these genes are indeed organized in two clusters, or whether the duplicated genes are serial duplications within a single Hox cluster.

However, there is some additional data that point to large-scale duplication of chromosomal fragments or even complete genomes in the spider. So far we also have found in our PCR screens several other genes that are present in two or more copies in the transcriptome of the spider Cupiennius, like extradenticle, homothorax, H15, Wnt5, Wnt7, engrailed, Delta, Suppressor of hairless, Krüppel, runt, pairberry, optomotor blind odd-skipped, apterous, orthodenticle [1931323334353637, our unpublished data]. In contrast, in most other arthropods most of these genes are present as one copy only. The relative high number of duplicated genes may point to a major duplication event in lineage to the spider, which might be caused by a whole genome duplication. A spider genome project would help to verify this.

Now data from all ten different arthropod Hox gene classes are known from this spider, another fact becomes obvious, that we already recognized previously based on a smaller data set 38, but which becomes even more prominent by new data on Cs-pb, Cs-Scr, and Cs-ftz [this paper, 19]. There are two discrete posterior expression boundaries for Hox genes in the Cupiennius (Fig 6). The expression of all anterior Hox genes (lab, pb, Hox3, Dfd, Scr, ftz) ends at the boundary between fourth walking leg (L4) and first opisthosomal segment (O1), which is at the tagma boundary between prosoma and opisthosoma. Also the posterior Hox genes (Antp, Ubx, abdA, AbdB) all have the same posterior expression border: the very posterior end of the embryo. There is only one Hox gene, Cs-Antp, that crosses the tagma boundary (Fig 6). In other arthropods, but also in vertebrates, most of these posterior expression borders are not defined as well as in the spider 1239.

The reason for the two discrete posterior expression borders remains unclear and we only can speculate on this. Between L4 and O1 is an important morphological boundary, the one between the two tagmata of the spider: the prosoma and the opisthosoma. The Hox genes might play a role in the specification of this boundary. In contrast, several Hox genes cross tagmata borders in other arthropods [e.g. 1223]. If the Hox genes play a role in tagma border specification, then this must be a peculiarity of the spider.

Another explanation could be that the anterior Hox genes are required for the specification of the different appendages in the spider. All six anterior Hox genes are expressed in distinct patterns within the appendages suggesting a role of them in appendage specification [151719, this paper] (see also Fig 3 and 4). It has been shown that Hox gene expression is associated with morphological diversification of leg segments in insects 40. Indications for interactions between Hox genes in the spider legs come also from the weaker expression of Cs-Dfd-2 in L3 and L4 that coincides with the stronger expression of Cs-Scr-1, Cs-Scr-2 and Cs-ftz in L3 and L4 (Fig. 3F, Fig. 4C, Fig. 4F and reference 19. Thus there might be a cross regulation between these Hox genes in the legs. Such a role of these Hox genes in the legs may form the reason for shared posterior expression boundaries (Fig. 6). The border between segments with and without appendages coincides with the tagma boundary between prosoma and opisthosoma. Spiders have true appendages on six segments: the cheliceres, the pedipalps, and four pair of walking legs. The more posterior segments do not have true appendages, however the second to fifth opisthosomal segment develop limb buds that give rise to the respiratory organs and the spinnerets 41.

A third possible explanation could be that the discrete Hox gene expression boundary is a result of the segmentation process that acts more upstream in the regulatory cascade and that lays down the segments. It is known from insects that the segmentation gene cascade indeed also controls the expression of Hox genes 424344. In insects this is mainly done by orthologs of gap genes. It is not known yet what genes regulate the expression of the spider Hox genes. A number of spider Hox genes obey parasegmental boundaries, as they do in Drosophila 31. Parasegmental boundaries are important developmental boundaries in the early embryo and are specified by the segmentation gene cascade 45. Segment-polarity genes like wingless, cubitus interruptus and engrailed maintain the parasegmental boundaries in arthropods. We assume that at least in part the same upstream acting regulatory machinery controls the segment-polarity genes and Hox genes in the spider as their expression boundaries match exactly. The discrete posterior expression border of the spider Hox genes is the result of genes that control them and these therefore may be an output of the upstream segmentation machinery that also control the expression boundaries of the segment-polarity genes. The assumption that the discrete Hox gene expression boundary could be the result of the segmentation process is strengthened by previous work in Cupiennius that suggested that there may be at least partially a difference in the mechanisms that specify the anterior segments and the posterior segments 34. The posterior segments form sequentially from a posterior growth zone and may be partially regulated in a different way. The discrete boundary of Hox gene expression at the prosoma-opisthosoma boundary therefore could reflect such a difference in the regulation of segmentation between the anterior and the posterior segments.

Embryos were obtained from our Cupiennius salei Keyserling (Chelicerata, Arachnida, Araneae, Ctenidae) colony in Cologne. Cocoons with embryos were taken from fertilized adult female spiders, RNA was isolated and reverse transcribed into cDNA as described previously 31, embryos were treated and fixed as described previously 1731.

Initial PCR fragments of the Cs-pb, Cs-Dfd-2, Cs-Scr-1, and Cs-Scr-2 were obtained using primer combinations directed against sequences in the homeodomain. We used the following primer combinations: for Cs-pb and Cs-Dfd-2 the primers 1521 (5'-GGA TTC TAY CCI TGG ATG-3') and 1520 (5'-CAT ICK ICK RTT YTG RAA CCA-3'), for Cs-Scr the primers 1289 (5'-CCN CAR ATH TAY CCN TGG ATG-3') and 1290 (5'-TT CCA YTT CAT NCG NCK RTT WTG- 3'). Additional sequences have been obtained by subsequent RACE-PCR. Additional sequence of the At-Dfd-1 gene has also been obtained via RACE-PCR, using primers based on a short sequence published by Abzhanov et al 14. The clones have been sequenced in both directions and the sequences are available under the accession numbers AM419029 to AM419032.

The expression patterns of the genes have been analyzed by in situ hybridizations using digoxigenin labelled anti-sense RNA probes 17. We used the following probes: for Cs-Dfd-1, Cs-Ubx-1 and Cs-Ubx-2 we used probes prepared from the cDNA clones isolated from the cDNA library 15 (accession numbers CAA07498, CAA07500, CAA07501), for Cs-pb, Cs-Scr-1 and Cs-Scr-2 we used probes prepared from the all available cDNA sequence (accession numbers AM419029, AM419030, AM419031), for Cs-Dfd-2 we used a probe prepared from a 3'RACE fragment, corresponding to nt 84–1407 of the available cDNA information (accession number AM419032).