There are seven Pax genes in Drosophila and nine Pax genes known in mouse and human. Different Pax proteins use multiple combinations of the HTH motifs to recognize several types of target sites. Drosophila Paired protein can bind, in vitro exclusively through its PAI domain (the N-terminal portion of the bipartite paired domain), or through a dimer of its Homeodomain, or through cooperative interaction between PAI domain and HD. However, paired function in vivo requires the synergistic action of both the PAI domain and the HD. Pax proteins with only a PD (such as Pax-5) appear to require both PAI and RED domains, while a Pax-6 isoform and a new Pax protein Lune, may rely on the RED domain and HD. Thus Pax protein appear to recognize different target genes in vivo through various combinations of their DNA binding domains, thus expanding their recognition repertoire (Jun, 1996).

Pax proteins play a diverse role in early animal development and contain the characteristic paired domain, consisting of two conserved helix-turn-helix motifs. In many Pax proteins the paired domain is fused to a second DNA binding domain of the paired-like homeobox family. By amino acid sequence alignments, secondary structure prediction, 3D-structure comparison, and phylogenetic reconstruction, the relationship between Pax proteins and members of the Tc1 family of transposases, which possibly share a common ancestor with Pax proteins, has been examined. It is suggested that the DNA binding domain of an ancestral transposase (proto-Pax transposase) was fused to a homeodomain shortly after the emergence of metazoans about one billion years ago. Using the transposase sequences as an outgroup the early evolution of the Pax proteins was examined. This novel evolutionary scenario features a single homeobox capturing event and an early duplication of Pax genes before the divergence of porifera, indicating a more diverse role of Pax proteins in primitive animals than previously expected (Breitling, 2000).

An attemp has been made to reconstruct the phylogeny and to reliably root the phylogenetic tree of Pax proteins. Since homeodomains, which have been compared for that purpose, are only present in some of the Pax proteins and are conspicuously absent in the PaxA/neuro and Pax1-9 group, the analysis was restricted to the paired box itself. This was facilitated by the introduction of a novel outgroup. Comparison of the X-ray structures of the paired box of Drosophila Paired (1PDN) and human Pax6 (6PAX) within the database of 3D-structures has revealed that the N-terminal subdomain (PAI domain) is closely related to the DNA binding domain of Tc3 transposase of Caenorhabditis elegans (1TC3). A general similarity between transposase DNA binding domains and the paired domain has been reported and their structural relationship has been observed during the analysis of the transposase structure. Initial Blast searches identified a group of transposases from C. elegans whose DNA binding domain seems to be more closely related to the paired box than to most other transposases. The DNA binding domain of these C. elegans transposases (proteins K03H6.3, W04G5.1, F26H9.3, F49C5.8, and C27H2.1; accession numbers T33011, T26169, T21438, T22423, and T19530) shows highly significant similarity only to Bmmar1, a transposase from Bombyx mori [accession number AAB47739, E-score (E)=2e-27 compared to K03H6.3], and to many Pax proteins (e.g. Hydra magnapapillata Pax2/5/8, E=9e-05; Phallusia mammilata Pax6 E=3e-04; or Paracentrotus lividus Pax1/9 E=6e-04). The DNA binding domains of other transposases yield E-scores worse than 1e-03 (e.g. Anopheles albimanus transposase AAB02109, E=9e-03). It is supposed that the transposases of C. elegans and B. mori might represent molecular fossils (proto-Pax) from the time before a homeobox capturing event took place, during which the catalytic domain of the transposase was lost and the DNA-binding domain was fused to a homeobox yielding the first PAX protein. If this is indeed the case, the proto-Pax transposases should also contain the C-terminal subdomain (RED domain) of the paired box. This subdomain is less conserved among Pax proteins than the PAI domain and does not show significant homology in sequence alignments between transposases and Pax proteins. A secondary structure analysis of the proto-Pax transposases was performed using a consensus method (Jpred2), which predicted that they indeed contain two helix-turn-helix motifs, homologous to both the PAI and the RED domain of Pax proteins (Breitling, 2000).

The observation that the DNA binding domain of transposases is in fact closely related to the paired box indicates that it should be possible to use them as an outgroup in the phylogenetic analysis of Pax proteins to determine the most likely evolutionary sequence. The transposase sequence (C. elegans K03H6.3, E = 2e-27) with the highest Blast score was compared to Pax proteins to generate a multiple sequence alignment of Pax-like transposases using the JPred2 server. The JPred2 algorithm was also used to generate a multiple sequence alignment for Pax proteins. Both alignments were combined and realigned by using ClustalW. The resulting data set contains transposases of the Tc1 and mariner families, as well as a wide range of Pax proteins from all known subgroups. The complete alignment was then used for phylogenetic analysis (Breitling, 2000).

Neighbor-joining and parsimony analysis reliably subdivides the Pax proteins into five large groups, which correspond to the classical subfamilies Pax1-9/Pax meso, PaxD/3-7/Gooseberry/Paired, PaxB/2-5-8/Sparkling, Pax4-6/Eyeless and PaxA/Pax neuro. The internal topology of the subfamilies agrees fairly well with the accepted evolutionary relationship of the organisms. One exception is the Pax4-6/Eyeless subfamily which is extremely conserved, so that an unambiguous determination of the internal branching order was not possible. The position of Drosophila Eyegone is also unreliable, because this protein contains only a partial paired domain. In both trees PaxC is significantly associated with the PaxA/Pax neuro subfamily, although PaxC carries a homeobox, and PaxA/Pax neuro proteins do not. Neighbor-joining and parsimony tree reconstruction place the Pax family within the Tc1 family of transposases, while it was not possible to identify a single closest relative of the paired box. The supposed proto-Pax transposases from C. elegans and B. mori, as identified by Blast searches, are not reliably placed as a sister-group of the Pax proteins. This might be due to the general difficulty of reconstructing well-resolved phylogenetic trees of the transposase family (Breitling, 2000).

This focus on the paired box as a descendant of a Tc1- like transposase DNA binding domain allowed for a reevaluation of the early evolution of the paired domain. These results show that the evolutionary scenario proposed by Galliot and Miller (2000) is unlikely to correctly represent the evolution of Pax proteins. This hypothesis was based mainly on the assumption that PaxA, which consists only of a paired box, resembles the probable ancestor of Pax proteins. Contrary to that idea, the scenario developed here is based on the assumption that the paired box is originally derived from a transposase and indicates that PaxA is probably derived by a secondary loss of the homeobox of a PaxC-like protein. These observations also make unlikely the hypothesis that there was more than one homeodomain capturing event. Furthermore, they suggest that the first duplication of Pax proteins occurred before the divergence of the porifera. This consequently implies that sponges, which lack nerve cells and most of the organs patterned by Pax genes in higher animals, already contained (at least) two Pax genes. The function of these early Pax proteins remains a mystery (Breitling, 2000).

Pair-rule genes were identified and named for their role in segmentation in embryos of the long germ insect Drosophila. Among short germ insects these genes exhibit variable expression patterns during segmentation and thus are likely to play divergent roles in this process. Understanding the details of this variation should shed light on the evolution of the genetic hierarchy responsible for segmentation in Drosophila and other insects. The expression of homologs of the Drosophila Pax group III genes paired, gooseberry and gooseberry-neuro have been examined in short germ flour beetles and grasshoppers. During Drosophila embryogenesis, paired acts as one of several pair-rule genes that define the boundaries of future parasegments and segments, via the regulation of segment polarity genes such as gooseberry, which in turn regulates gooseberry-neuro, a gene expressed later in the developing nervous system. Using a crossreactive antibody, it has been shown that the embryonic expression of Pax group III genes in both the flour beetle Tribolium and the grasshopper Schistocerca is remarkably similar to the pattern in Drosophila. Two Pax group III genes, pairberry1 and pairberry2, are responsible for the observed protein pattern in grasshopper embryos. Both pairberry1 and pairberry2 are expressed in coincident stripes of a one-segment periodicity, in a manner reminiscent of Drosophila gooseberry and gooseberry-neuro. pairberry1, however, is also expressed in stripes of a two-segment periodicity before maturing into its segmental pattern. This early expression of pairberry1 is reminiscent of Drosophila paired and represents the first evidence for pair-rule patterning in short germ grasshoppers or any hemimetabolous insect (Davis, 2001).

With the exception of possible nematode homologs, protostome Pax group III (PgIII) genes have thus far not been reported outside Drosophila. Two PgIII genes from Schistocerca have been named pairberry1 (pby1) and pairberry2 (pby2). Each gene possesses both a paired box and an extended S₅₀ paired-like homeobox. Phylogenetic analysis and high sequence similarity to Drosophila prd, gsb and gsbn supports the inclusion of pby1 and pby2 within PgIII. Additionally, pby1 and pby2 appear to be more closely related to each other than either is to prd, gsb or gsbn, suggesting they may be the result of an independent duplication. This conclusion is tempered, however, by the possibility of homogenization of pby1 and pby2 via gene conversion (Davis, 2001).

Although the two grasshopper genes may be closely related, their relationship to the fly genes could not be unequivocally resolved. Although it is possible that pby1 and pby2 result from the duplication of the ancestral gsb/gsbn gene along the lineage leading to Schistocerca after its split with Drosophila, this scenario implies that a grasshopper prd ortholog either exists and has not been found, or was subsequently lost. Since the expression patterns of both pby1 and pby2 include elements similar to the expression of each of the three Drosophila genes, it is thought more likely that pby1 and pby2 result from an independent duplication of a single ancestral insect PgIII (prd/gsb/gsbn) gene (Davis, 2001).

The early transcript and protein expression patterns of pby1 provide, for the first time, evidence of pair-rule patterning in the grasshopper Schistocerca. Indirect evidence is provided by the order of appearance of the gnathal and thoracic Pby1 stripes. In particular, the onset of the Mx and T1 stripes is delayed relative to their adjacent stripes. Thus, like many segment polarity genes in Drosophila, the order of appearance of these segmental stripes follows a two-segment periodicity. This may reflect, as it does in Drosophila, regulation by an underlying pair-rule patterning mechanism (Davis, 2001).

Stronger evidence for pair-rule patterning lies with the early domains of pby1 expression from T2 to A10. Stripes of these segments originate as broad domains of a two-segment periodicity at the extending posterior tip, each of which subsequently splits into a pair of adjacent segmental stripes. Thus, adjacent stripes arise by subtly different means. The segmental stripes of T2, A1, A3, A5, A7 and A9 resolve from the anterior edge of sequentially appearing broad domains. By contrast, the segmental stripes of T3, A2, A4, A6, A8 and A10 resolve from within the posterior portions of the same respective broad domains. This resolution of broad domains into adjacent pairs of segmental stripes is analogous to the process by which Drosophila prd acquires its segmental pattern from initial stripes of a two-segment periodicity (Davis, 2001).

Although similar to Drosophila and flour beetles, the broad domains in grasshopper exhibit at least one notable difference. When compared with either Drosophila or flour beetles, the pairing of stripes in grasshoppers is shifted by one segment. For example, in grasshoppers, the Pby1 stripes of A1 and A2 derive from a single A1/A2 broad domain, while in flies and flour beetles the segmental Pby stripes of A1 and A2 derive from the T3/A1 and A2/A3 broad domains (primary Prd stripes 4 and 5 in Drosophila). The shift in phasing of stripe pairs in Schistocerca when compared with Tribolium and Drosophila is reflected in the fact that the grasshopper A11 Pby1 stripe, which appears relatively late, arises without a sister stripe. Such variation in phasing is likely to reflect a spatial shift in the expression of upstream components of the segmentation hierarchy (Davis, 2001).

An additional similarity of early pby1 expression to Drosophila prd is its timing relative to segment polarity genes. In Drosophila, prd is expressed before en and wg. In Schistocerca, pby1 is expressed before En protein by approximately four to five stripes from ~20%-27% of development. Hence, pby1 is also likely to be expressed ahead of wg. As in Schistocerca gregaria, wg transcript appears only two to three stripes ahead of En protein. Another feature shared by the early pby1 pattern and that of Drosophila prd is the gnathal arc. This early domain comprises the future Pby1 stripes of the Mn, Mx and La segments. In Drosophila, prd is also expressed as a single broad stripe before splitting into primary Prd stripes 1 and 2 at the onset of cellularization, just as stripes 3-7 begin to appear. Primary stripes 1 and 2 in turn give rise to the future Mn, Mx and La secondary stripes of Prd. This early Prd domain in flies is thus remarkably similar to the Pby1 gnathal arc in grasshoppers. A similar pattern in flour beetle embryos could not be detected, since the Mn Pby stripe appears de novo (Davis, 2001).

The position of Pby1 stripes just anterior to En with an overlap of ~one cell row, along with their subsequent restriction to the neuroectoderm, is reminiscent of gsb expression in Drosophila. Similarly, the delayed appearance of Pby2 stripes, their restricted form, and their coincident expression with Pby1 anterior to En is reminiscent of late gsb expression. Additionally, the striped neural expression of both pby1 and pby2 as late as 40% of development is reminiscent of gsbn expression. Thus, only one of two PgIII genes identified in Schistocerca, pby1, is potentially functioning in the capacity of all three PgIII genes in Drosophila (prd, gsb and gsbn), while pby2 is potentially functioning in the capacity of one, or perhaps two, of the Drosophila genes (gsb and gsb-n). Finally, although the behavior pby2 is most similar to Drosophila gsb and gsbn, the late expression of both pby1 and pby2 at the base of the developing gnathal appendages is reminiscent of the late expression of prd at the base of the gnathal protuberances in Drosophila embryos (Davis, 2001).

During Drosophila embryogenesis, the pair-rule gene prd activates the segment polarity gene gsb, which, in turn, activates gsbn. Additionally, the products of these three genes are for the most part functionally interchangeable. Given both their similarity to the three fly genes and their coincident expression, pby1 may be required for the activation of pby2 (Davis, 2001).

In Drosophila, prd is also required for the activation of odd-numbered wg stripes. Thus, Pby1 may be required for activation of wg in Schistocerca americana. The temporal dynamics of wg mRNA in the closely related grasshopper Schistocerca gregaria are consistent with this suggestion. In Drosophila, prd is also responsible for activating and defining the posterior border of odd-numbered En stripes. This is suggested by the absence of odd-numbered En stripes in prd-negative embryos, as well as their posterior expansion in heat shocked prd embryos. Consistent with this role, the posterior borders of secondary Prd stripes in Drosophila are coincident with the posterior borders of En stripes. In Schistocerca, however, Pby1 does not simultaneously share a posterior border with En. Instead, nascent segmental stripes spanning four cell rows narrow to two cell rows just before the appearance of an adjacent En stripe, which overlaps by only a single row of cells. This lack of temporally coincident expression does not, however, rule out a possible role in activating en, for it is conceivable that the four-cell row domain of Pby1 may activate en before narrowing, with the result that En appears specifically in cells that were previously expressing pby1. A similar situation may hold true for Drosophila, as it has been proposed that, despite the coincident expression of secondary Prd stripes and En, it is instead the earlier primary stripes of Prd that are responsible for the activation of en. Finally, it is important to note that a fully functioning pair-rule mechanism in grasshoppers may well require genes in addition to pby1 that exhibit pair-rule like expression patterns (Davis, 2001).

Based on widespread conservation of expression patterns, it seems likely that the Drosophila segment polarity genes functioned as such in the context of the ancestral insect segmentation system. The picture is less clear for pair-rule genes. In light of the more basal phylogenetic position of Schistocerca, it is tempting to view the posterior expression domains of eve and ftz as ancestral for insects, existing before the evolutionary recruitment of these genes to play a role in segmentation. In support of this conjecture, vertebrate orthologs of eve are linked to the Hox clusters and expressed in broad Hox-like domains, while the C. elegans eve ortholog, vab-7, is both expressed in a broad posterior domain and required for posterior cell fates. ftz, a gene closely related to the Antp-class Hox genes, is likewise expressed in a broad Hox-like domain in mites (Davis, 2001).

However, grasshoppers in some respects are likely to represent a derived state for insects. This is probably the case for eve, since this gene is expressed in stripes in spiders. Thus, it is possible that eve was primitively expressed in both stripes and a posterior domain, but somewhere along the lineage leading to Schistocerca, the gene lost its striped expression. The observation that a PgIII gene is expressed in stripes of a two-segment periodicity in grasshoppers suggests that pair-rule patterning is part of the ancestral insect segmentation system. However, confirmation of this claim will require closer examination of the striped expression of pair-rule orthologs in primitive insects and non-insect arthropods (Davis, 2001).

An additional consequence of the molecular data presented here is that Tribolium and Schistocerca appear more similar in their embryology than previously appreciated. Before this study, the non-striped expression of eve and ftz did not allow comparison with the striped expression of pair-rule genes in other insects. The Pby pattern, however, allows such a comparison. In the case of Tribolium, only one Pby stripe, that of the mandibular segment, has formed before the onset of gastrulation; eve and ftz stripes at this stage have likewise not formed posterior to the gnathal region. In Schistocerca, no pby1 expression has been detected before the onset of gastrulation (~36 hours AEL), and the first stripe associated with segmentation (the gnathal arc) does not appear until 10% of develoment (~50 hours AEL), well after gastrulation has begun. Thus, neither Tribolium nor Schistocerca has specified segmental or parasegmental boundaries posterior to the head at the start of gastrulation, conforming to the classical idea of short (as opposed to intermediate) germ embryogenesis (Davis, 2001).

Drosophila prd is at the bottom of the genetic hierarchy of pair-rule genes and this fact, coupled with its later segmental expression, have led some to suggest that in flies prd acts as a bridge between the pair-rule and segment polarity levels of the segmentation hierarchy. If pair-rule patterning is an evolutionarily recent specialization of prd, then the segmental secondary Prd stripes of Drosophila are best seen as the remnants of an ancestral dual function as a pair-rule and segment polarity gene. It is perhaps not surprising then, that pby1 -- a PgIII gene from a more phylogenetically primitive insect -- is expressed in both a pair-rule and segment polarity fashion. As one of only two PgIII genes in Schistocerca, pby1 is expressed in a manner reminiscent of the combined pattern of all three PgIII genes in Drosophila. In lacking the specialized expression of the Drosophila genes, pby1 may be the closest approximation of the ancestral insect PgIII gene (Davis, 2001).