The Trichoplax PaxB Gene: A Putative Proto-PaxA/B/C Gene Predating the Origin of Nerve and Sensory Cells
http://www.100md.com
分子生物学进展 2005年第7期
* ITZ, Ecology & Evolution, Hannover, Germany; Department of Invertebrate Zoology & Molecular Labs, American Museum of Natural History; Department of Physiology and Neuroscience, NYU Medical Center; and Yale University, MCDB
Correspondence: E-mail: thorsten.hadrys@uni-oldenburg.de.
Abstract
Pax genes play key regulatory roles in embryonic and sensory organ development in metazoans but their evolution and ancestral functions remain widely unresolved. We have isolated a Pax gene from Placozoa, beside Porifera the only metazoan phylum that completely lacks nerve and sensory cells or organs. These simplest known metazoans also lack any kind of symmetry, organs, extracellular matrix, basal lamina, muscle cells, and main body axis. The isolated Pax gene from Trichoplax adhaerens harbors a paired domain, an octapeptide, and a full-length homeodomain. It displays structural features not only of PaxB and Pax2/5/8-like genes but also of PaxC and Pax6 genes. Conserved splice sites between Placozoa, Cnidaria, and triploblasts, mark the ancient origin of intron structures. Phylogenetic analyses demonstrate that the Trichoplax PaxB gene, TriPaxB, is basal not only to all other known PaxB genes but also to PaxA and PaxC genes and their relatives in triploblasts (namely Pax2/5/8, Pax4/6, and Poxneuro). TriPaxB is expressed in distinct cell patches near the outer edge of the animal body, where undifferentiated and possibly multipotent cells are found. This expression pattern indicates a developmental role in cell-type specification and/or differentiation, probably in specifying-determining fiber cells, which are regarded as proto-neural/muscle cells in Trichoplax. While PaxB, Pax2/5/8, and Pax6 genes have been linked to nerve cell and sensory system/organ development in virtually all animals investigated so far, our study suggests that Pax genes predate the origin of nerve and sensory cells.
Key Words: PaxB ? Pax gene evolution ? Proto-Pax ? Trichoplax ? Placozoa
Introduction
Transcription factors of the Pax gene family serve crucial functions in several developmental processes, particularly with respect to the development and differentiation of the central nervous system and sensory organs, both in vertebrates and invertebrates (Walther et al. 1991; Halder, Callaerts, and Gehring 1995; Rinkwitz-Brandt, Arnold, and Bober 1996; Torres, Gomez-Pardo, and Gruss 1996; Callaerts et al. 1999; Czerny et al. 1999; Holland et al. 1999; Kavaler et al. 1999; Kozmik et al. 1999; Groger et al. 2000; Kozmik et al. 2003). A structural characteristic of Pax genes is a paired-type DNA-binding domain, which was first identified in the Drosophila pair-rule gene paired (Frigerio et al. 1986). In addition, most Pax genes contain a complete or partial homeodomain, and some Pax genes also possess an octapeptide close to the C-terminal of the paired domain. Regions between domains are less well conserved. Based on their structural properties Pax genes are grouped into five subfamilies in triploblasts: namely Pax1-9/Poxmeso, Pax2-5-8/sparkling, Pax3-7/paired/gooseberry, Pax6-4/eyeless, and Poxneuro (Breitling and Gerber 2000; Miller et al. 2000). In diploblasts, Pax genes from Cnidaria and Porifera belong to four classes, PaxA–D (Balczarek, Lai, and Kumar 1997; Sun et al. 1997; Catmull et al. 1998; Hoshiyama et al. 1998; Groger et al. 2000; Miller et al. 2000; Kozmik et al. 2003). Orthological relationships between diploblast and tripoblast Pax genes are not finally resolved yet, but it has been proposed that cnidarian PaxB genes are related to triploblast Pax2/5/8 genes, PaxC genes to Pax4-6, and PaxD genes to Pax1-9 and Pax3-7 genes (Sun et al. 1997; Miller et al. 2000).
Pax2/5/8-related genes seem to be primarily associated with the development of mechanosensory systems in both invertebrates and vertebrates. In higher vertebrates Pax2, Pax5, and Pax8 genes are present in multiple paralogs (likely as a result of chromosomal or whole-genome duplications) and are expressed in the developing inner ear and central nervous system in mammals (Rinkwitz-Brandt, Arnold, and Bober 1996; Torres, Gomez-Pardo, and Gruss 1996). In Drosophila only one Pax2/5/8 gene (D-Pax2; sparkling) is present, which has a crucial function for the development of mechanosensory bristles (Fu et al. 1998; Kavaler et al. 1999), ommatidial cone, and pigment cells (Fu and Noll 1997). A Pax2/5/8 gene identified in ascidians is expressed in the atrial primordium (Wada et al. 1998), a structure that comprises sensory cells similar to those of the vertebrate inner ear (Bone 1978), and in gastropods, a Pax2/5/8 gene is expressed in the statocyst (O'Brien and Degnan 2003). Interestingly the cnidarian Pax2/5/8 counterpart, PaxB, is implicated in nerve cell differentiation in a hydrozoan (Groger et al. 2000) and in sensory organ (statocyst and eye) development in the cubozoan Tripedalia cystophora (Kozmik et al. 2003). The latter mirrors the combined expression (and function) of Pax6 (eye) and Pax2/5/8 (statocyst) genes in triploblastic animals. The Tripedalia Pax gene, TcPaxB, not only unites functional but also structural features of Pax2/5/8 and Pax6-like genes. The paired domain is similar to Pax2/5/8 genes, whereas the homeodomain displays features of Pax6-like genes. Kozmik et al. (2003) demonstrated that the PaxB protein is a functional hybrid of Pax2/5/8 and Pax6.
Different hypotheses on the origin of metazoan Pax genes have been proposed. One hypothesis suggests that a PaxA-like paired domain was fused to a homeodomain and founded the Pax gene family (Galliot and Miller 2000; Miller et al. 2000). Breitling and Gerber (2000) postulated that Pax-like genes evolved by fusion of a DNA-binding domain of an ancestral transposase (Proto-Pax transposase) to a homeodomain shortly after the emergence of metazoan animals about 1 billion years ago. The authors further propose a single homeodomain fusion event followed by an early duplication of Pax genes before the divergence of Porifera. In order to unravel the early evolution of Pax genes we need data from all putative basal metazoan groups. While Pax genes have been isolated from sponges and cnidarians, no data have been available from the last and possibly most crucial diploblast phylum, the Placozoa.
Here we report the isolation and characterization of a single Pax gene from the morphologically most simple organized metazoan animal, the placozoan Trichoplax adhaerens, which lacks any kind of nervous system and/or sensory organs. It is important to note that Placozoa are not secondarily reduced cnidarians (Ender and Schierwater 2003), and thus lack of nerve cells most likely is a plesiomorphy. The Trichoplax Pax gene, TriPaxB, is expressed in distinct cell patches in a ring-shaped pattern near the lower-upper epithelium boundary.
Our structural and phylogenetic analyses show that the Trichoplax Pax gene is basal to PaxA-, B- and C-type genes and harbors structural features of both Pax2/5/8 and Pax6 genes. These findings suggest that TriPaxB gave rise to at least four of the five Pax gene families in higher metazoan animals and provide support for Millers' hypothesis on the origin of Pax genes (Miller et al. 2000). The TriPaxB gene meets expectations for a Proto-Pax gene or the early descendant of a Proto-Pax gene in metazoan animals.
Materials and Methods
Polymerase Chain Reaction Amplification of Paired-Box Sequences
Trichoplax genomic DNA was isolated as described previously (Ender and Schierwater 2003). Messenger RNA (mRNA) from growing and reproducing Trichoplax individuals was isolated using the Invitrogen (San Diego, Calif.) "Micro-Fast Track" Kit according to the manufacturer's protocol. Different sets of degenerate primers were used to amplify a 344-bp fragment of the paired domain. Two sets of degenerate primers are described in Hoshiyama et al. (1998). A third set was designed based upon conserved paired domain sequences from other diploblastic Pax genes by using the CODEHOP program (http://blocks.fhcrc.org/blocks/codehop.html). Complementary DNA (cDNA) preparations from growing and reproducing Trichoplax individuals served as template DNA. Double-stranded cDNA was synthesized from mRNA using the "Creator Smart" System (Clontech, Palo Alto, Calif.) according to the manufacturer's protocol. Polymerase chain reaction (PCR) fragments were obtained only with the primer pairs S1-AS3 and S2-AS3; the sequences are as follows (see also Hoshiyama et al. 1998): forward: S1 5'-CAGGATCCCARYTIGGNGGNGTNTT- (corresponding to the "QLGGVF" motif); reverse: AS3 5'-GTGAATTCATYTCCCANGCRAADAT-3' (corresponding to IFAWEI). Other primers designed from highly conserved amino acid sequences within the paired domain did not result in the amplification of Pax-specific PCR products. These primers were forward: S2 5'-GAGGATCCTTYGTNAAYGGNMGNCC-3' (corresponding to FVNGRP); reverse: AS1 5'-GTGAATTCYKRTCNKDATYTCCCA-3' (corresponding to WEIRD[RK]); see Hoshiyama et al. (1998), and two primers designed using CODEHOP: forward: S3 5'-CAAGATCCTGTGCCGGTACTAYGARACNGG-3' (corresponding to KILSRYYETG) and reverse: 5'-CTCCAGCAGGCAGTCCCKDATYTCCCA-3' (corresponding to WEIRDCLLQ). PCR conditions were 30 s 95°C, 30 s 50°C, and 60 s 68°C, and 40 cycles were performed. PCR fragments were subcloned in pGEM-T vector Promega, Madison, Wisc.). Plasmid minipreparations were sequenced in both directions using ABI (Foster City, Calif.) BigDye terminator chemistry on an ABI-310 capillar sequencer.
Rapid Amplification of cDNA Ends and Genome Walk PCR
Starting from the paired-box cDNA fragment, the coding sequence of Trichoplax PaxB was amplified using the "SMART RACE" system (Clontech). The following primers were designed from the sequence of the isolated paired-box cDNA fragment (forward: 3' Walk 1: ATCAACTACCGTTGGTGTTGCCACCT; 3' Walk 2: CGATATGACGACGTATTGCTTCACGC; reverse: 5' Walk 1: CTTGCTTCCTCCAATAATACCTGGGC; 5' Walk 2: CTTCCATTTTCAAACACACCACCCAG). The 3' and 5' rapid amplification of cDNA ends (RACE)-PCR reactions were performed according to the manufacturer's manual (Clontech). PCR conditions were 95°C 15 s, 68°C 3 min, 35 cycles. The obtained RACE products were subcloned (pGEM-T) and sequenced.
To characterize the corresponding Trichoplax PaxB gene structure a "Genome Walk" (Clontech) was carried out. PCR reactions were performed using long-template Taq-polymerase as described in the manufacturer's manual. PCR fragments were subcloned (pGEM-T) and characterized by sequencing.
Expression Analyses
Whole-mount in situ hybridization experiments were performed using a modified protocol developed for Cnidaria and Placozoa, respectively (Groger et al. 2000; Jakob, et al. 2004). Animals were fixed in Lavdowsky's fixative as described in Jakob et al. (2004). TriPaxB-, Trox2-, and actin-RNA-Probes were synthesized from subcloned cDNA fragments (pGEM-T easy; Promega) using digoxigenin (DIG) and fluorescein isothiocyanate–uridine triphosphate (FITC-UTP) labeling (Roche, Mannheim, Germany) according to the manufacturer's manual. Reverse transcriptase (RT)–PCR was done as described previously (Hadrys et al. 2004).
Phylogenetic Analyses
Distance and Maximum Parsimony analyses were carried out in order to infer phylogenetic relationships between Pax genes. All known paired domain sequences from diploblasts were included in the analysis. For rooted tree analyses a Pseudomonas transposase sequence served as an out-group (Breitling and Gerber 2000).
Bayes analysis was done with MrBayes (Huelsenbeck and Ronquist 2001). The parsmodel was applied and the following parameters were used: Markov chain Monte Carlo (MCMC) with 100,000 four chains and sampling frequency of 10. The trees generated from the MCMC simulation were imported into PAUP, and a Bayesian tree was visualized using the 50% majority rule option in PAUP (Swofford 2003).
For likelihood ratio tests the method of Shimodaira and Hasegawa (1999, 2001) as implemented in the PROML program in PHYLIP (Felsenstein 2003, http://evolution.genetics.washington.edu/phylip.html) was used to calculate likelihood ratios of the best neighbor-joining (NJ) tree and the parsimony tree relative to trees that were generated by RETREE (PHYLIP package) that removed and regrafted the TriPaxB gene at all nodes in the parsimony tree.
In this way the basal position of the TriPaxB gene could be tested in comparison to its position within the PaxA–C clade and within the PaxB clade (see fig. 3). The likelihoods of over 20 trees generated by RETREE were included in these tests using a likelihood that took into account site-specific rate differences using a gamma correction.
FIG. 3.— (A) Rooted neighbor-joining tree of paired domain sequences of all known diploblast PaxA, PaxB, PaxC, and PaxD genes. Bootstrap values result from 1,000 stepwise addition replicates. The shown topology is identical to the parsimony tree. A pseudomonas transposase sequence serves as out-group. (B) Bayesian analysis for all known diploblast Pax genes. The Bayes proportions are shown on the branches. At 95% Bayesian proportion, a basal position of the TriPaxB gene compared to PaxA–C and PaxB clades is supported. In this figure the structural features of the corresponding Pax genes are also illustrated. Paired domains = light gray boxes, homeodomains = black boxes, octapetides = white circles. For CqPaxB so far only sequences of the paired domain are available. EfPax258 contains a partial homeodomain. (C) Neighbor-joining tree of all known diploblast together with several triploblast Pax gene paired domains. Note that the inclusion of triploblast sequences results in a loss of bootstrap support. Am, Acropora millepora; Cc, Cladonema californicum; Cq, Chrysaora quinquecirrha; Dm, Drosophila melanogaster; Ef, Ephydatia fluviatilis; Hl, Hydra littoralis; Hr, Halocynthia roretzi; Mm, Mus musculus; Pc, Podocoryne carnea; Pl, Paracentrotus lividus; Tc, Tripedalia cystophora; Tri, Trichoplax adhaerens.
Accession numbers for sequences included in the analyses are Acropora millepora PaxA (AmPaxA): AF053458 [GenBank] ; A. millepora PaxC (AmPaxC): AF053459 [GenBank] ; A. millepora PaxD (AmPaxD): AF241311 [GenBank] ; Chrysaora quinquecirrha PaxA1 (CqPaxA1): U96195 [GenBank] ; Cladonema californicum PaxB (CcPaxB): AF260128 [GenBank] ; Chrysaora quinquecirrha PaxB (CqPaxB): U96197 [GenBank] ; Drosophila melanogaster eyeless (ey): X79493 [GenBank] ; Drosophila melanogaster paired (prd): M14548 [GenBank] ; Ephydatia fluviatilis Pax2/5/8 (EfPax258): AB007462 [GenBank] ; Halocynthia roretzi Pax-37 (HrPax-37): D84254 [GenBank] ; Hydra littoralis PaxA (HlPaxA): U96194 [GenBank] ; Hydra littoralis PaxB (HlPaxB): U96194 [GenBank] ; Mus Musculus Pax2: A60086 [GenBank] ; Mus musculus Pax3: NM_008781 [GenBank] ; Mus Musculus Pax5: M97013 [GenBank] ; Mus musculus Pax6: BC011272 [GenBank] ; Paracentrotus lividus Pax258 (suPax258): AF016884 [GenBank] ; Podocoryne carnea (PcPaxB): AJ249563 [GenBank] ; Pseudomonas syringae transposase: AF169828 [GenBank] ; T. cystophora PaxB (TcPaxB): AY280703 [GenBank] .
Results
Isolation and Structural Features of the T. adhaerens PaxB Gene
Using different sets of primers, we obtained PCR fragments of the expected size (344 bp) with the primer combination S1/AS3 only. Because primers were designed according to the most conserved regions of the paired-box motif, they are expected to amplify paired-box sequences of all Pax gene subfamilies. From a total of 20 clones sequenced all of which were 100% identical in sequence. By means of 5' and 3' RACE 955 nucleotides of the coding sequence, including the paired and homeodomain (318 amino acids) were isolated (fig. 1). The 3'RACE reactions also revealed the presence of two weaker, slightly larger PCR products, indicating the presence at least two alternative transcripts (data not shown).
FIG. 1.— Nucleotide and amino acid sequence of TriPaxB. The isolated coding region comprises 955 nucleotides (318 amino acids). The TriPaxB protein includes a paired domain (underlayed in gray), an octapeptide (boxed) and a "full" homeodomain (underlayed in light gray). The positions of two introns are indicated by arrowheads (black). An intron located directly upstream the paired domain was found in all Pax genes investigated so far. An intron located between the octapeptide and the homeodomain was also found in other Pax2/5/8 genes, for example, in Drosophila D-Pax2 (sparkling; here the intron is much longer, however).
The Trichoplax Pax gene contains a paired domain, an octapeptide, and a "full-length" homeodomain (figs. 1 and 2). The paired domain displays structural features of PaxB and Pax2/5/8 genes, and it harbors several amino acid positions that are regarded as diagnostic for this class of proteins (Kozmik et al. 2003; fig. 1). The full-length homeodomain, however, is more similar to PaxC/Pax6-like genes (fig. 2B).
FIG. 2.— Alignment of paired domain (A) and homeodomain (B) sequences of several PaxA, PaxB, PaxC, Pax2/5/8, and Pax6 genes. Amino acids underlayed in gray are Pax6 specific. Trichoplax PaxB and Tripedalia PaxB harbor paired domains that are Pax2/5/8 related and homeodomains that are PaxC/Pax6 related. Am: Acropora millepora; Dm: Drosophila melanogaster; Ef: Ephydatia fluviatilis; eye: eyeless; Hl: Hydra littoralis; Mm: Mus musculus; Pc: Podocoryne carnea; spar: sparkling; Tc: Tripedalia cystophora; Tri: Trichoplax adhaerens.
Two exon-intron junctions were mapped via genome walk PCR. The first intron is located directly upstream and adjacent to the paired box (fig. 2A). The location of this first intron is conserved in all Pax genes investigated so far. The second intron is located upstream of the homeodomain and comprises 350 bp. The accession number of the coding sequence is DQ22561.
Phylogenetic Analyses
In phylogenetic analyses the TriPaxB paired domain clusters basal to the PaxB/Pax2/5/8 subfamily (fig. 3). Furthermore, TriPaxB appears to be basal also relative to all but one Pax family. TriPaxB always comes out basal to PaxA, PaxB, and PaxC genes, independent of the algorithm and also independent of whether paired domain sequences from triploblastic animals were included or not. The topology shown in figure 3A does not change when randomly chosen paired domain sequences from triploblasts are added to the analysis (fig. 3C).
To test the robustness of the basal position of the TriPaxB gene relative to PaxA, PaxB, and PaxC genes, we used the likelihood ratio test as developed by Shimodaira and Hasegawa (1999, 2001). The results of tests using the PROML program in PHYLIP (Felsenstein 2003) indicated that the tree with TriPaxB placed basal to all other Pax genes (except for PaxD) was the best tree according to likelihood scores (table 1). Furthermore, any tree tested where the TriPaxB gene was placed in the PaxA clade was highly statistically significantly indicated as worse than the TriPaxB basal tree. Placement of the TriPaxB gene into the clade in figure 3A that holds most of the other PaxB genes, however, indicates that while these trees have worse likelihood scores than the TriPaxB basal tree, the trees are not statistically significantly worse. In table 1, tree 1 is the "TriPaxB basal" tree. Trees 2–5 and 12 and 13 are trees where TriPaxB was grafted onto a PaxA or PaxC branch in the tree in figure 3A. All other trees except for tree 15 are cases where TriPaxB was grafted into places in the PaxB clade in figure 3A. Tree 15 retained TriPaxB as basal but as sister to the single PaxD gene.
Table 1 Results of Likelihood Ratio Tests (as Developed by Shimodaira and Hasegawa, 1999, 2001) Using the PROML Program in PHYLIP (Felsenstein, 2003)
A second approach we took was to examine the support for the NJ tree and the parsimony tree using Bayesian statistics. The Bayesian analysis suggests that the TriPaxB gene is not supported as a member of either the PaxA–C or PaxB clades and supports at 95% Bayesian proportion, the basal position of the TriPaxB gene. The Bayes proportions are shown on the branches of the tree in figure 3B.
Expression of TriPaxB
Semiquantitative RT-PCR experiments revealed that TriPaxB is expressed in adult, i.e., growing, and vegetatively reproducing animals. Here, TriPaxB expression is significantly higher than that of the regulatory Antp superclass gene, EMX but lower than expression of the HSP70 gene (fig. 5). Whole-mount in situ hybridization studies revealed expression in distinct cell patches along a ring region close to the outer edge of the animal body (fig. 6A, E, and F). Control hybridization with an actin antisense probe shows homogeneous expression throughout the entire body, as expected for a housekeeping gene (Fig. 6B). Control experiments with sense probes did not reveal any specific hybridization signals (data not shown, but see Jakob et al. 2004).
FIG. 5.— Semiquantitative RT-PCR analysis of HSP70, TriPaxB, and EMX using a cDNA preparation from reproducing stages of Trichoplax adhaerens. TriPaxB expression is significantly higher than expression of the regulatory Antp superclass gene, EMX, but lower than expression of the HSP70 gene.
FIG. 6.— Whole-mount in situ expression analyses of TriPaxB, Trox2, and actin. The TriPaxB gene is expressed in growing and vegetatively reproducing stages of Trichoplax adhaerens (cf. fig. 5) in cell patches at some distance to the outer edge of the animal body (A, C–F). Actin (B) and Trox2 (G) expression is shown for comparison. A–D represent DIG-labeled whole-mount in situ hybridizations, whereas E–G were labeled with FITC-UTP. Note that TriPaxB is not expressed in differentiated fiber cells that are located in between the upper and the lower epithelium in the center of the animal body (C and D). Trox2 (G) is also expressed in the proposed zone of cell proliferation and differentiation close the outer edge of the animal body, but on average stronger and more homogeneously (Jakob et al. 2004).
Interestingly, expression signals found in smaller animals were weaker than those found in larger animals (compare fig. 6A and E; data not shown). Analysis of tissue sections revealed that TriPaxB-expressing cells are not epithelial cells but cells inside the animal (fig. 6C and D). Possibly these cells are undifferentiated fiber cells. Interestingly, the Hox/ParaHox gene, Trox2, shows a similar spatial expression pattern in this region of cell differentiation (on average the Trox2 signal, however, is stronger and more evenly spaced; fig. 6G and Jakob et al. 2004).
Discussion
The structure of the putative ancestral Pax gene has been controversially discussed (cf. Sun et al. 1997; Catmull et al. 1998; Hoshiyama et al. 1998; Breitling and Gerber 2000). The data obtained from T. adhaerens strongly suggest that a putative "Proto-Pax" gene harbored at least two of the three Pax gene motifs, that is, a paired domain and a homeodomain, and possibly also the third motif, the octapeptide (fig. 4). The transition from an "Ur-Pax" gene (a homeodomain fused to a paired domain; Breitling and Gerber 2000; Miller et al. 2000) to the Trichoplax PaxABC gene, requires only a single step, the incorporation of the octapeptide. From this "fully loaded" ancestral PaxABC gene other Pax genes may be derived by subsequent deletion events, that is, loss of the (1) homeodomain, (2) partial homeodomain, and/or (3) octapeptide (fig. 4). The supported model is consistent with the paired domain gene trees in figure 3, current knowledge on the phylogenetic position of Placozoa, and the comparison of Pax gene functions.
FIG. 4.— Pax gene evolution model. A Proto-Pax gene (B or D like) derived from a gene fusion event between a Proto-Pax transposase and a homeodomain (HD) in protozoans, as first proposed by Breitling and Gerber (2000). The octapeptide (OP) capturing occurred either after or before the first gene duplication event. In the latter case the octapeptide got lost in the D lineage. Two more rounds of gene duplication (B > C and C > A) followed by partial losses of homeodomain and/or octapeptide sequences did lead to the current Pax gene assembly in cnidarians. In sponges a partial loss of the homeodomain took place. Proposed relationships to triploblast Pax258, Pax6, and Pax37 genes are indicated by dotted lines. For abbreviations see figure 3. Paired domains = green boxes, homeodomains = black boxes, octapetides = blue circles.
The Pax evolution model in figure 4 incorporates the assumption that a Proto-Pax gene derived from a gene fusion event between a paired domain (e.g., from a Proto-Pax transposase) (Breitling and Gerber 2000) and a homeodomain in protozoans. In addition to the original arguments, this scenario seems plausible also because no Pax genes have been found in protists and the best-supported Pax paired domain phylogeny is obtained with transposase as out-group. The diploblast Pax gene tree (fig. 3A and B) is in accordance with a proposed basal position of Placozoa (for overview and references see Syed and Schierwater 2002). If Porifera were basal, however, the Pax evolution model would require a slight modification. Here, the last common ancestor of Placozoa and Porifera would be assumed to have harbored a Trichoplax-type PaxB gene, whose structure remained unchanged in Placozoa but experienced a partial loss of the homeodomain in the lineage leading to Porifera. Because it is known that Placozoa are not secondarily reduced Cnidaria (Ender and Schierwater 2003) and because their bauplan cannot easily be derived from a sponge bauplan (Syed and Schierwater 2002), one has to argue that Pax genes of the A class, and—particularly interesting—also of the B and C class, predated the invention of nerve and sensory cells.
With respect to the evolution of PaxD-like genes additional research is needed to decide whether PaxD branched off even earlier (as shown in our evolution model) or if the PaxD paired domain originated in the next common ancestor of Cnidaria and triploblasts. In the first scenario a PaxD-type gene either got lost or escaped surveys in Placozoa and Porifera. In the second scenario insufficient taxon sampling or nonoptimal out-group choice may have hindered phylogenetic resolution in the analysis. Although it seems unlikely that a PaxD-type gene escaped our PCR screen, we cannot rule out, however, that Placozoans posses more than one Pax gene. To decide between the alternatives more data will be needed, which will likely come from ongoing whole-genome sequencing efforts in Placozoa, Porifera, and Cnidaria.
Functional information from triploblast Pax genes may also add to our understanding of early duplication events in diploblastic animals. It was previously assumed that PaxB is a precursor of Pax2/5/8, whereas PaxC could be a precursor of Pax6 genes in triploblasts. Plaza et al. (2003) recently demonstrated that DNA-binding characteristics of cnidarian PaxB and PaxC proteins display no simple relationship to Pax2/5/8 and Pax6 genes. The authors showed that A. millepora PaxB and PaxC proteins can both bind to eyeless (ey) targets in vivo and in vitro, which casts doubt on the postulated direct relationship between cnidarian Pax genes and the bilaterian Pax6 and Pax2/5/8 classes. Given that our analysis suggests that Cnidarian PaxB and PaxC (and also PaxA) genes are derived from a gene with similar organization to the placozoan PaxB-like gene, one could speculate that the ancestral PaxABC gene unites functional features which were retained in cnidarian PaxB and PaxC as well as in triploblast Pax2/5/8 and Pax6 genes. This hypothesis is indeed supported by Kozmik et al. (2003) who showed that T. cystophora PaxB (the only Pax gene in cubomedusa found so far) contains a Pax2/5/8-type paired domain and octapeptide but a Pax6-type homeodomain. The Tripedalia PaxB gene is expressed in larval stages, in the retina, lens, and statocyst. According to functional properties, that is, binding specificity, the ability to rescue spa (a Drosphila Pax2 mutant) and to induce ectopic eyes in Drosophila, the authors suggest that the ancestor of the cubozoan PaxB-like protein was the primordial Pax protein in eye evolution and that Pax6-like genes evolved in triploblasts after separation from Cnidaria. Trichoplax PaxB meets these expectations for an ancestral Proto-PaxABC gene. Most interestingly, Trichoplax does not possess any kind of sensory organs or nerve cells. Expression of TriPaxB in small irregular cell patches along the outer edge of the animal possibly relates to undifferentiated cells and is spatially overlapping with the Trox2 expression domain (the only Hox/ParaHox gene found in Placozoa; see fig. 6G and Jakob et al. 2004). Quite noteworthy, TriPaxB is not expressed in differentiated fiber cells, which represent putative proto-neural/muscular cells and are located between the upper and lower epithelium throughout the center region of the body (fig. 6C and D). TriPaxB could, however, function in cell determination of fiber cells from undifferentiated and multipotent precursor cells (cf. Jakob et al. 2004). We propose that TriPaxB and Trox2 both demark a particular zone of cell proliferation and differentiation.
We further propose that PaxB was co-opted in the last common ancestor of cnidarians and triploblasts for sensory organ and nerve cell development and that two rounds of gene duplication (B C and C A) followed by partial losses of homeodomain and/or octapeptide sequences led to the current Pax gene assembly in cnidarians (fig. 4). In the triploblast lineage additional duplication-deletion events have taken place and among others resulted in the functional split of protein function of Pax2/5/8 and Pax6 genes (fig. 4). Because TriPaxB is basal to all other known PaxB genes (and also to PaxA and PaxC genes), it is basal also to Pax2/5/8 and Pax6 genes (Sun et al. 1997; fig. 4). Our data suggest that a PaxB similar gene (harboring functional features of both Pax2/5/8 and Pax6 genes) was the original gene involved in sensory organ development and evolution. A functional split into Pax2/5/8 (mechanosensory) and Pax6 (eye/light sense) likely occurred in the last common ancestor of diploblasts and triploblasts. While the placozoan TriPaxB gene most likely predates the origin of nerve and sensory cells, its ancestral developmental function needs to be investigated in more detail.
Acknowledgements
We acknowledge very helpful comments from two anonymous reviewers and support from the Deutsche Forschungsgemeinschaft (DFG Schi 277/10) and Human Frontier Science Program (HFSP RGP0221/2001-M).
References
Balczarek, K. A., Z. C. Lai, and S. Kumar. 1997. Evolution of functional diversification of the paired box (Pax) DNA-binding domains. Mol. Biol. Evol. 14:829–842.
Bone, Q., and K. P. Ryan. 1978. Cupular sense organs in ciona (Tunicata: Ascidiacae). J. Zool. Lond. 186:417–429.
Breitling, R., and J. K. Gerber. 2000. Origin of the paired domain. Dev. Genes Evol. 210:644–650.
Callaerts, P., A. M. Munoz-Marmol, S. Glardon, E. Castillo, H. Sun, W. H. Li, W. J. Gehring, and E. Salo. 1999. Isolation and expression of a Pax-6 gene in the regenerating and intact Planarian Dugesia(G)tigrina. Proc. Natl. Acad. Sci. USA 96:558–563.
Catmull, J., D. C. Hayward, N. E. McIntyre, J. S. Reece-Hoyes, R. Mastro, P. Callaerts, E. E. Ball, and D. J. Miller. 1998. Pax-6 origins—implications from the structure of two coral pax genes. Dev. Genes Evol. 208:352–356.
Czerny, T., G. Halder, U. Kloter, A. Souabni, W. J. Gehring, and M. Busslinger. 1999. Twin of eyeless, a second Pax-6 gene of Drosophila, acts upstream of eyeless in the control of eye development. Mol. Cell 3:297–307.
Ender, A., and B. Schierwater. 2003. Placozoa are not derived cnidarians: evidence from molecular morphology. Mol. Biol. Evol. 20:130–134.
Felsenstein, J. 2003. Inferring Phylogenies. Sinauer Associates, Inc., Sunderland, Mass.
Frigerio, G., M. Burri, D. Bopp, S. Baumgartner, and M. Noll. 1986. Structure of the segmentation gene paired and the Drosophila PRD gene set as part of a gene network. Cell 47:735–746.
Fu, W., H. Duan, E. Frei, and M. Noll. 1998. Shaven and sparkling are mutations in separate enhancers of the Drosophila Pax2 homolog. Development 125:2943–2950.
Fu, W., and M. Noll. 1997. The Pax2 homolog sparkling is required for development of cone and pigment cells in the Drosophila eye. Genes Dev. 11:2066–2078.
Galliot, B., and D. Miller. 2000. Origin of anterior patterning. How old is our head? Trends Genet. 16:1–5.
Groger, H., P. Callaerts, W. J. Gehring, and V. Schmid. 2000. Characterization and expression analysis of an ancestor-type Pax gene in the hydrozoan jellyfish Podocoryne carnea. Mech. Dev. 94:157–169.
Hadrys, T., V. Prince, M. Hunter, R. Baker, and S. Rinkwitz. 2004. Comparative genomic analysis of vertebrate Hox3 and Hox4 genes. J. Exp. Zool. B Mol. Dev. Evol. 302:147–164.
Halder, G., P. Callaerts, and W. J. Gehring. 1995. Induction of ectopic eyes by targeted expression of the eyeless gene in Drosophila. Science 267:1788–1792.
Holland, L. Z., M. Schubert, Z. Kozmik, and N. D. Holland. 1999. AmphiPax3/7, an amphioxus paired box gene: insights into chordate myogenesis, neurogenesis, and the possible evolutionary precursor of definitive vertebrate neural crest. Evol. Dev. 1:153–165.
Hoshiyama, D., H. Suga, N. Iwabe, M. Koyanagi, N. Nikoh, K. Kuma, F. Matsuda, T. Honjo, and T. Miyata. 1998. Sponge Pax cDNA related to Pax-2/5/8 and ancient gene duplications in the Pax family. J. Mol. Evol. 47:640–648.
Huelsenbeck, J. P., and F. Ronquist. 2001. MRBAYES: Bayesian inference of phylogenetic trees. Bioinformatics 17:754–755.
Jakob, W., S. Sagasser, S. Dellaporta, P. Holland, K. Kuhn, and B. Schierwater. 2004. The Trox-2 Hox/ParaHox gene of Trichoplax (Placozoa) marks an epithelial boundary. Dev. Genes Evol. 214:170–175.
Kavaler, J., W. Fu, H. Duan, M. Noll, and J. W. Posakony. 1999. An essential role for the Drosophila Pax2 homolog in the differentiation of adult sensory organs. Development 126:2261–2272.
Kozmik, Z., M. Daube, E. Frei, B. Norman, L. Kos, L. J. Dishaw, M. Noll, and J. Piatigorsky. 2003. Role of Pax genes in eye evolution: a cnidarian PaxB gene uniting Pax2 and Pax6 functions. Dev. Cell 5:773–785.
Kozmik, Z., N. D. Holland, A. Kalousova, J. Paces, M. Schubert, and L. Z. Holland. 1999. Characterization of an amphioxus paired box gene, AmphiPax2/5/8: developmental expression patterns in optic support cells, nephridium, thyroid-like structures and pharyngeal gill slits, but not in the midbrain-hindbrain boundary region. Development 126:1295–1304.
Miller, D. J., D. C. Hayward, J. S. Reece-Hoyes, I. Scholten, J. Catmull, W. J. Gehring, P. Callaerts, J. E. Larsen, and E. E. Ball. 2000. Pax gene diversity in the basal cnidarian Acropora millepora (Cnidaria, Anthozoa): implications for the evolution of the Pax gene family. Proc. Natl. Acad. Sci. USA 97:4475–4480.
O'Brien, E. K., and B. M. Degnan. 2003. Expression of Pax258 in the gastropod statocyst: insights into the antiquity of metazoan geosensory organs. Evol. Dev. 5:572–578.
Plaza, S., D. M. De Jong, W. J. Gehring, and D. J. Miller. 2003. DNA-binding characteristics of cnidarian Pax-C and Pax-B proteins in vivo and in vitro: no simple relationship with the Pax-6 and Pax-2/5/8 classes. J. Exp. Zool. B Mol. Dev. Evol. 299:26–35.
Rinkwitz-Brandt, S., H. H. Arnold, and E. Bober. 1996. Regionalized expression of Nkx5-1, Nkx5-2, Pax2 and sek genes during mouse inner ear development. Hear Res. 99:129–138.
Shimodaira, H., and M. Hasegawa. 1999. Multiple comparisons of log-likelihoods with applications to phylogenetic inference. Mol. Biol. Evol. 16:1114–1116.
———. 2001. CONSEL: for assessing the confidence of phylogenetic tree selection. Bioinformatics 17:1246–1247.
Sun, H., A. Rodin, Y. Zhou, D. P. Dickinson, D. E. Harper, D. Hewett-Emmett, and W. H. Li. 1997. Evolution of paired domains: isolation and sequencing of jellyfish and hydra Pax genes related to Pax-5 and Pax-6. Proc. Natl. Acad. Sci. USA 94:5156–5161.
Swofford, D. L. 2003. PAUP*: phylogenetic analysis using parsimony (*and other methods). Version 4. Sinauer Associates Inc., Sunderland, Mass.
Syed, T., and B. Schierwater 2002. Trichoplax adhaerens: discovered as a missing link, forgotten as a hydrozoan, re-discovered as a key to metazoan evolution. Vie Milieu 52:177–187.
Torres, M., E. Gomez-Pardo, and P. Gruss. 1996. Pax2 contributes to inner ear patterning and optic nerve trajectory. Development 122:3381–3391.
Wada, H., H. Saiga, N. Satoh, and P. W. Holland. 1998. Tripartite organization of the ancestral chordate brain and the antiquity of placodes: insights from ascidian Pax-2/5/8, Hox and Otx genes. Development 125:1113–1122.
Walther, C., J. L. Guenet, D. Simon, U. Deutsch, B. Jostes, M. D. Goulding, D. Plachov, R. Balling, and P. Gruss. 1991. Pax: a murine multigene family of paired box-containing genes. Genomics 11:424–434.(Thorsten Hadrys*,,,1, Rob)
Correspondence: E-mail: thorsten.hadrys@uni-oldenburg.de.
Abstract
Pax genes play key regulatory roles in embryonic and sensory organ development in metazoans but their evolution and ancestral functions remain widely unresolved. We have isolated a Pax gene from Placozoa, beside Porifera the only metazoan phylum that completely lacks nerve and sensory cells or organs. These simplest known metazoans also lack any kind of symmetry, organs, extracellular matrix, basal lamina, muscle cells, and main body axis. The isolated Pax gene from Trichoplax adhaerens harbors a paired domain, an octapeptide, and a full-length homeodomain. It displays structural features not only of PaxB and Pax2/5/8-like genes but also of PaxC and Pax6 genes. Conserved splice sites between Placozoa, Cnidaria, and triploblasts, mark the ancient origin of intron structures. Phylogenetic analyses demonstrate that the Trichoplax PaxB gene, TriPaxB, is basal not only to all other known PaxB genes but also to PaxA and PaxC genes and their relatives in triploblasts (namely Pax2/5/8, Pax4/6, and Poxneuro). TriPaxB is expressed in distinct cell patches near the outer edge of the animal body, where undifferentiated and possibly multipotent cells are found. This expression pattern indicates a developmental role in cell-type specification and/or differentiation, probably in specifying-determining fiber cells, which are regarded as proto-neural/muscle cells in Trichoplax. While PaxB, Pax2/5/8, and Pax6 genes have been linked to nerve cell and sensory system/organ development in virtually all animals investigated so far, our study suggests that Pax genes predate the origin of nerve and sensory cells.
Key Words: PaxB ? Pax gene evolution ? Proto-Pax ? Trichoplax ? Placozoa
Introduction
Transcription factors of the Pax gene family serve crucial functions in several developmental processes, particularly with respect to the development and differentiation of the central nervous system and sensory organs, both in vertebrates and invertebrates (Walther et al. 1991; Halder, Callaerts, and Gehring 1995; Rinkwitz-Brandt, Arnold, and Bober 1996; Torres, Gomez-Pardo, and Gruss 1996; Callaerts et al. 1999; Czerny et al. 1999; Holland et al. 1999; Kavaler et al. 1999; Kozmik et al. 1999; Groger et al. 2000; Kozmik et al. 2003). A structural characteristic of Pax genes is a paired-type DNA-binding domain, which was first identified in the Drosophila pair-rule gene paired (Frigerio et al. 1986). In addition, most Pax genes contain a complete or partial homeodomain, and some Pax genes also possess an octapeptide close to the C-terminal of the paired domain. Regions between domains are less well conserved. Based on their structural properties Pax genes are grouped into five subfamilies in triploblasts: namely Pax1-9/Poxmeso, Pax2-5-8/sparkling, Pax3-7/paired/gooseberry, Pax6-4/eyeless, and Poxneuro (Breitling and Gerber 2000; Miller et al. 2000). In diploblasts, Pax genes from Cnidaria and Porifera belong to four classes, PaxA–D (Balczarek, Lai, and Kumar 1997; Sun et al. 1997; Catmull et al. 1998; Hoshiyama et al. 1998; Groger et al. 2000; Miller et al. 2000; Kozmik et al. 2003). Orthological relationships between diploblast and tripoblast Pax genes are not finally resolved yet, but it has been proposed that cnidarian PaxB genes are related to triploblast Pax2/5/8 genes, PaxC genes to Pax4-6, and PaxD genes to Pax1-9 and Pax3-7 genes (Sun et al. 1997; Miller et al. 2000).
Pax2/5/8-related genes seem to be primarily associated with the development of mechanosensory systems in both invertebrates and vertebrates. In higher vertebrates Pax2, Pax5, and Pax8 genes are present in multiple paralogs (likely as a result of chromosomal or whole-genome duplications) and are expressed in the developing inner ear and central nervous system in mammals (Rinkwitz-Brandt, Arnold, and Bober 1996; Torres, Gomez-Pardo, and Gruss 1996). In Drosophila only one Pax2/5/8 gene (D-Pax2; sparkling) is present, which has a crucial function for the development of mechanosensory bristles (Fu et al. 1998; Kavaler et al. 1999), ommatidial cone, and pigment cells (Fu and Noll 1997). A Pax2/5/8 gene identified in ascidians is expressed in the atrial primordium (Wada et al. 1998), a structure that comprises sensory cells similar to those of the vertebrate inner ear (Bone 1978), and in gastropods, a Pax2/5/8 gene is expressed in the statocyst (O'Brien and Degnan 2003). Interestingly the cnidarian Pax2/5/8 counterpart, PaxB, is implicated in nerve cell differentiation in a hydrozoan (Groger et al. 2000) and in sensory organ (statocyst and eye) development in the cubozoan Tripedalia cystophora (Kozmik et al. 2003). The latter mirrors the combined expression (and function) of Pax6 (eye) and Pax2/5/8 (statocyst) genes in triploblastic animals. The Tripedalia Pax gene, TcPaxB, not only unites functional but also structural features of Pax2/5/8 and Pax6-like genes. The paired domain is similar to Pax2/5/8 genes, whereas the homeodomain displays features of Pax6-like genes. Kozmik et al. (2003) demonstrated that the PaxB protein is a functional hybrid of Pax2/5/8 and Pax6.
Different hypotheses on the origin of metazoan Pax genes have been proposed. One hypothesis suggests that a PaxA-like paired domain was fused to a homeodomain and founded the Pax gene family (Galliot and Miller 2000; Miller et al. 2000). Breitling and Gerber (2000) postulated that Pax-like genes evolved by fusion of a DNA-binding domain of an ancestral transposase (Proto-Pax transposase) to a homeodomain shortly after the emergence of metazoan animals about 1 billion years ago. The authors further propose a single homeodomain fusion event followed by an early duplication of Pax genes before the divergence of Porifera. In order to unravel the early evolution of Pax genes we need data from all putative basal metazoan groups. While Pax genes have been isolated from sponges and cnidarians, no data have been available from the last and possibly most crucial diploblast phylum, the Placozoa.
Here we report the isolation and characterization of a single Pax gene from the morphologically most simple organized metazoan animal, the placozoan Trichoplax adhaerens, which lacks any kind of nervous system and/or sensory organs. It is important to note that Placozoa are not secondarily reduced cnidarians (Ender and Schierwater 2003), and thus lack of nerve cells most likely is a plesiomorphy. The Trichoplax Pax gene, TriPaxB, is expressed in distinct cell patches in a ring-shaped pattern near the lower-upper epithelium boundary.
Our structural and phylogenetic analyses show that the Trichoplax Pax gene is basal to PaxA-, B- and C-type genes and harbors structural features of both Pax2/5/8 and Pax6 genes. These findings suggest that TriPaxB gave rise to at least four of the five Pax gene families in higher metazoan animals and provide support for Millers' hypothesis on the origin of Pax genes (Miller et al. 2000). The TriPaxB gene meets expectations for a Proto-Pax gene or the early descendant of a Proto-Pax gene in metazoan animals.
Materials and Methods
Polymerase Chain Reaction Amplification of Paired-Box Sequences
Trichoplax genomic DNA was isolated as described previously (Ender and Schierwater 2003). Messenger RNA (mRNA) from growing and reproducing Trichoplax individuals was isolated using the Invitrogen (San Diego, Calif.) "Micro-Fast Track" Kit according to the manufacturer's protocol. Different sets of degenerate primers were used to amplify a 344-bp fragment of the paired domain. Two sets of degenerate primers are described in Hoshiyama et al. (1998). A third set was designed based upon conserved paired domain sequences from other diploblastic Pax genes by using the CODEHOP program (http://blocks.fhcrc.org/blocks/codehop.html). Complementary DNA (cDNA) preparations from growing and reproducing Trichoplax individuals served as template DNA. Double-stranded cDNA was synthesized from mRNA using the "Creator Smart" System (Clontech, Palo Alto, Calif.) according to the manufacturer's protocol. Polymerase chain reaction (PCR) fragments were obtained only with the primer pairs S1-AS3 and S2-AS3; the sequences are as follows (see also Hoshiyama et al. 1998): forward: S1 5'-CAGGATCCCARYTIGGNGGNGTNTT- (corresponding to the "QLGGVF" motif); reverse: AS3 5'-GTGAATTCATYTCCCANGCRAADAT-3' (corresponding to IFAWEI). Other primers designed from highly conserved amino acid sequences within the paired domain did not result in the amplification of Pax-specific PCR products. These primers were forward: S2 5'-GAGGATCCTTYGTNAAYGGNMGNCC-3' (corresponding to FVNGRP); reverse: AS1 5'-GTGAATTCYKRTCNKDATYTCCCA-3' (corresponding to WEIRD[RK]); see Hoshiyama et al. (1998), and two primers designed using CODEHOP: forward: S3 5'-CAAGATCCTGTGCCGGTACTAYGARACNGG-3' (corresponding to KILSRYYETG) and reverse: 5'-CTCCAGCAGGCAGTCCCKDATYTCCCA-3' (corresponding to WEIRDCLLQ). PCR conditions were 30 s 95°C, 30 s 50°C, and 60 s 68°C, and 40 cycles were performed. PCR fragments were subcloned in pGEM-T vector Promega, Madison, Wisc.). Plasmid minipreparations were sequenced in both directions using ABI (Foster City, Calif.) BigDye terminator chemistry on an ABI-310 capillar sequencer.
Rapid Amplification of cDNA Ends and Genome Walk PCR
Starting from the paired-box cDNA fragment, the coding sequence of Trichoplax PaxB was amplified using the "SMART RACE" system (Clontech). The following primers were designed from the sequence of the isolated paired-box cDNA fragment (forward: 3' Walk 1: ATCAACTACCGTTGGTGTTGCCACCT; 3' Walk 2: CGATATGACGACGTATTGCTTCACGC; reverse: 5' Walk 1: CTTGCTTCCTCCAATAATACCTGGGC; 5' Walk 2: CTTCCATTTTCAAACACACCACCCAG). The 3' and 5' rapid amplification of cDNA ends (RACE)-PCR reactions were performed according to the manufacturer's manual (Clontech). PCR conditions were 95°C 15 s, 68°C 3 min, 35 cycles. The obtained RACE products were subcloned (pGEM-T) and sequenced.
To characterize the corresponding Trichoplax PaxB gene structure a "Genome Walk" (Clontech) was carried out. PCR reactions were performed using long-template Taq-polymerase as described in the manufacturer's manual. PCR fragments were subcloned (pGEM-T) and characterized by sequencing.
Expression Analyses
Whole-mount in situ hybridization experiments were performed using a modified protocol developed for Cnidaria and Placozoa, respectively (Groger et al. 2000; Jakob, et al. 2004). Animals were fixed in Lavdowsky's fixative as described in Jakob et al. (2004). TriPaxB-, Trox2-, and actin-RNA-Probes were synthesized from subcloned cDNA fragments (pGEM-T easy; Promega) using digoxigenin (DIG) and fluorescein isothiocyanate–uridine triphosphate (FITC-UTP) labeling (Roche, Mannheim, Germany) according to the manufacturer's manual. Reverse transcriptase (RT)–PCR was done as described previously (Hadrys et al. 2004).
Phylogenetic Analyses
Distance and Maximum Parsimony analyses were carried out in order to infer phylogenetic relationships between Pax genes. All known paired domain sequences from diploblasts were included in the analysis. For rooted tree analyses a Pseudomonas transposase sequence served as an out-group (Breitling and Gerber 2000).
Bayes analysis was done with MrBayes (Huelsenbeck and Ronquist 2001). The parsmodel was applied and the following parameters were used: Markov chain Monte Carlo (MCMC) with 100,000 four chains and sampling frequency of 10. The trees generated from the MCMC simulation were imported into PAUP, and a Bayesian tree was visualized using the 50% majority rule option in PAUP (Swofford 2003).
For likelihood ratio tests the method of Shimodaira and Hasegawa (1999, 2001) as implemented in the PROML program in PHYLIP (Felsenstein 2003, http://evolution.genetics.washington.edu/phylip.html) was used to calculate likelihood ratios of the best neighbor-joining (NJ) tree and the parsimony tree relative to trees that were generated by RETREE (PHYLIP package) that removed and regrafted the TriPaxB gene at all nodes in the parsimony tree.
In this way the basal position of the TriPaxB gene could be tested in comparison to its position within the PaxA–C clade and within the PaxB clade (see fig. 3). The likelihoods of over 20 trees generated by RETREE were included in these tests using a likelihood that took into account site-specific rate differences using a gamma correction.
FIG. 3.— (A) Rooted neighbor-joining tree of paired domain sequences of all known diploblast PaxA, PaxB, PaxC, and PaxD genes. Bootstrap values result from 1,000 stepwise addition replicates. The shown topology is identical to the parsimony tree. A pseudomonas transposase sequence serves as out-group. (B) Bayesian analysis for all known diploblast Pax genes. The Bayes proportions are shown on the branches. At 95% Bayesian proportion, a basal position of the TriPaxB gene compared to PaxA–C and PaxB clades is supported. In this figure the structural features of the corresponding Pax genes are also illustrated. Paired domains = light gray boxes, homeodomains = black boxes, octapetides = white circles. For CqPaxB so far only sequences of the paired domain are available. EfPax258 contains a partial homeodomain. (C) Neighbor-joining tree of all known diploblast together with several triploblast Pax gene paired domains. Note that the inclusion of triploblast sequences results in a loss of bootstrap support. Am, Acropora millepora; Cc, Cladonema californicum; Cq, Chrysaora quinquecirrha; Dm, Drosophila melanogaster; Ef, Ephydatia fluviatilis; Hl, Hydra littoralis; Hr, Halocynthia roretzi; Mm, Mus musculus; Pc, Podocoryne carnea; Pl, Paracentrotus lividus; Tc, Tripedalia cystophora; Tri, Trichoplax adhaerens.
Accession numbers for sequences included in the analyses are Acropora millepora PaxA (AmPaxA): AF053458 [GenBank] ; A. millepora PaxC (AmPaxC): AF053459 [GenBank] ; A. millepora PaxD (AmPaxD): AF241311 [GenBank] ; Chrysaora quinquecirrha PaxA1 (CqPaxA1): U96195 [GenBank] ; Cladonema californicum PaxB (CcPaxB): AF260128 [GenBank] ; Chrysaora quinquecirrha PaxB (CqPaxB): U96197 [GenBank] ; Drosophila melanogaster eyeless (ey): X79493 [GenBank] ; Drosophila melanogaster paired (prd): M14548 [GenBank] ; Ephydatia fluviatilis Pax2/5/8 (EfPax258): AB007462 [GenBank] ; Halocynthia roretzi Pax-37 (HrPax-37): D84254 [GenBank] ; Hydra littoralis PaxA (HlPaxA): U96194 [GenBank] ; Hydra littoralis PaxB (HlPaxB): U96194 [GenBank] ; Mus Musculus Pax2: A60086 [GenBank] ; Mus musculus Pax3: NM_008781 [GenBank] ; Mus Musculus Pax5: M97013 [GenBank] ; Mus musculus Pax6: BC011272 [GenBank] ; Paracentrotus lividus Pax258 (suPax258): AF016884 [GenBank] ; Podocoryne carnea (PcPaxB): AJ249563 [GenBank] ; Pseudomonas syringae transposase: AF169828 [GenBank] ; T. cystophora PaxB (TcPaxB): AY280703 [GenBank] .
Results
Isolation and Structural Features of the T. adhaerens PaxB Gene
Using different sets of primers, we obtained PCR fragments of the expected size (344 bp) with the primer combination S1/AS3 only. Because primers were designed according to the most conserved regions of the paired-box motif, they are expected to amplify paired-box sequences of all Pax gene subfamilies. From a total of 20 clones sequenced all of which were 100% identical in sequence. By means of 5' and 3' RACE 955 nucleotides of the coding sequence, including the paired and homeodomain (318 amino acids) were isolated (fig. 1). The 3'RACE reactions also revealed the presence of two weaker, slightly larger PCR products, indicating the presence at least two alternative transcripts (data not shown).
FIG. 1.— Nucleotide and amino acid sequence of TriPaxB. The isolated coding region comprises 955 nucleotides (318 amino acids). The TriPaxB protein includes a paired domain (underlayed in gray), an octapeptide (boxed) and a "full" homeodomain (underlayed in light gray). The positions of two introns are indicated by arrowheads (black). An intron located directly upstream the paired domain was found in all Pax genes investigated so far. An intron located between the octapeptide and the homeodomain was also found in other Pax2/5/8 genes, for example, in Drosophila D-Pax2 (sparkling; here the intron is much longer, however).
The Trichoplax Pax gene contains a paired domain, an octapeptide, and a "full-length" homeodomain (figs. 1 and 2). The paired domain displays structural features of PaxB and Pax2/5/8 genes, and it harbors several amino acid positions that are regarded as diagnostic for this class of proteins (Kozmik et al. 2003; fig. 1). The full-length homeodomain, however, is more similar to PaxC/Pax6-like genes (fig. 2B).
FIG. 2.— Alignment of paired domain (A) and homeodomain (B) sequences of several PaxA, PaxB, PaxC, Pax2/5/8, and Pax6 genes. Amino acids underlayed in gray are Pax6 specific. Trichoplax PaxB and Tripedalia PaxB harbor paired domains that are Pax2/5/8 related and homeodomains that are PaxC/Pax6 related. Am: Acropora millepora; Dm: Drosophila melanogaster; Ef: Ephydatia fluviatilis; eye: eyeless; Hl: Hydra littoralis; Mm: Mus musculus; Pc: Podocoryne carnea; spar: sparkling; Tc: Tripedalia cystophora; Tri: Trichoplax adhaerens.
Two exon-intron junctions were mapped via genome walk PCR. The first intron is located directly upstream and adjacent to the paired box (fig. 2A). The location of this first intron is conserved in all Pax genes investigated so far. The second intron is located upstream of the homeodomain and comprises 350 bp. The accession number of the coding sequence is DQ22561.
Phylogenetic Analyses
In phylogenetic analyses the TriPaxB paired domain clusters basal to the PaxB/Pax2/5/8 subfamily (fig. 3). Furthermore, TriPaxB appears to be basal also relative to all but one Pax family. TriPaxB always comes out basal to PaxA, PaxB, and PaxC genes, independent of the algorithm and also independent of whether paired domain sequences from triploblastic animals were included or not. The topology shown in figure 3A does not change when randomly chosen paired domain sequences from triploblasts are added to the analysis (fig. 3C).
To test the robustness of the basal position of the TriPaxB gene relative to PaxA, PaxB, and PaxC genes, we used the likelihood ratio test as developed by Shimodaira and Hasegawa (1999, 2001). The results of tests using the PROML program in PHYLIP (Felsenstein 2003) indicated that the tree with TriPaxB placed basal to all other Pax genes (except for PaxD) was the best tree according to likelihood scores (table 1). Furthermore, any tree tested where the TriPaxB gene was placed in the PaxA clade was highly statistically significantly indicated as worse than the TriPaxB basal tree. Placement of the TriPaxB gene into the clade in figure 3A that holds most of the other PaxB genes, however, indicates that while these trees have worse likelihood scores than the TriPaxB basal tree, the trees are not statistically significantly worse. In table 1, tree 1 is the "TriPaxB basal" tree. Trees 2–5 and 12 and 13 are trees where TriPaxB was grafted onto a PaxA or PaxC branch in the tree in figure 3A. All other trees except for tree 15 are cases where TriPaxB was grafted into places in the PaxB clade in figure 3A. Tree 15 retained TriPaxB as basal but as sister to the single PaxD gene.
Table 1 Results of Likelihood Ratio Tests (as Developed by Shimodaira and Hasegawa, 1999, 2001) Using the PROML Program in PHYLIP (Felsenstein, 2003)
A second approach we took was to examine the support for the NJ tree and the parsimony tree using Bayesian statistics. The Bayesian analysis suggests that the TriPaxB gene is not supported as a member of either the PaxA–C or PaxB clades and supports at 95% Bayesian proportion, the basal position of the TriPaxB gene. The Bayes proportions are shown on the branches of the tree in figure 3B.
Expression of TriPaxB
Semiquantitative RT-PCR experiments revealed that TriPaxB is expressed in adult, i.e., growing, and vegetatively reproducing animals. Here, TriPaxB expression is significantly higher than that of the regulatory Antp superclass gene, EMX but lower than expression of the HSP70 gene (fig. 5). Whole-mount in situ hybridization studies revealed expression in distinct cell patches along a ring region close to the outer edge of the animal body (fig. 6A, E, and F). Control hybridization with an actin antisense probe shows homogeneous expression throughout the entire body, as expected for a housekeeping gene (Fig. 6B). Control experiments with sense probes did not reveal any specific hybridization signals (data not shown, but see Jakob et al. 2004).
FIG. 5.— Semiquantitative RT-PCR analysis of HSP70, TriPaxB, and EMX using a cDNA preparation from reproducing stages of Trichoplax adhaerens. TriPaxB expression is significantly higher than expression of the regulatory Antp superclass gene, EMX, but lower than expression of the HSP70 gene.
FIG. 6.— Whole-mount in situ expression analyses of TriPaxB, Trox2, and actin. The TriPaxB gene is expressed in growing and vegetatively reproducing stages of Trichoplax adhaerens (cf. fig. 5) in cell patches at some distance to the outer edge of the animal body (A, C–F). Actin (B) and Trox2 (G) expression is shown for comparison. A–D represent DIG-labeled whole-mount in situ hybridizations, whereas E–G were labeled with FITC-UTP. Note that TriPaxB is not expressed in differentiated fiber cells that are located in between the upper and the lower epithelium in the center of the animal body (C and D). Trox2 (G) is also expressed in the proposed zone of cell proliferation and differentiation close the outer edge of the animal body, but on average stronger and more homogeneously (Jakob et al. 2004).
Interestingly, expression signals found in smaller animals were weaker than those found in larger animals (compare fig. 6A and E; data not shown). Analysis of tissue sections revealed that TriPaxB-expressing cells are not epithelial cells but cells inside the animal (fig. 6C and D). Possibly these cells are undifferentiated fiber cells. Interestingly, the Hox/ParaHox gene, Trox2, shows a similar spatial expression pattern in this region of cell differentiation (on average the Trox2 signal, however, is stronger and more evenly spaced; fig. 6G and Jakob et al. 2004).
Discussion
The structure of the putative ancestral Pax gene has been controversially discussed (cf. Sun et al. 1997; Catmull et al. 1998; Hoshiyama et al. 1998; Breitling and Gerber 2000). The data obtained from T. adhaerens strongly suggest that a putative "Proto-Pax" gene harbored at least two of the three Pax gene motifs, that is, a paired domain and a homeodomain, and possibly also the third motif, the octapeptide (fig. 4). The transition from an "Ur-Pax" gene (a homeodomain fused to a paired domain; Breitling and Gerber 2000; Miller et al. 2000) to the Trichoplax PaxABC gene, requires only a single step, the incorporation of the octapeptide. From this "fully loaded" ancestral PaxABC gene other Pax genes may be derived by subsequent deletion events, that is, loss of the (1) homeodomain, (2) partial homeodomain, and/or (3) octapeptide (fig. 4). The supported model is consistent with the paired domain gene trees in figure 3, current knowledge on the phylogenetic position of Placozoa, and the comparison of Pax gene functions.
FIG. 4.— Pax gene evolution model. A Proto-Pax gene (B or D like) derived from a gene fusion event between a Proto-Pax transposase and a homeodomain (HD) in protozoans, as first proposed by Breitling and Gerber (2000). The octapeptide (OP) capturing occurred either after or before the first gene duplication event. In the latter case the octapeptide got lost in the D lineage. Two more rounds of gene duplication (B > C and C > A) followed by partial losses of homeodomain and/or octapeptide sequences did lead to the current Pax gene assembly in cnidarians. In sponges a partial loss of the homeodomain took place. Proposed relationships to triploblast Pax258, Pax6, and Pax37 genes are indicated by dotted lines. For abbreviations see figure 3. Paired domains = green boxes, homeodomains = black boxes, octapetides = blue circles.
The Pax evolution model in figure 4 incorporates the assumption that a Proto-Pax gene derived from a gene fusion event between a paired domain (e.g., from a Proto-Pax transposase) (Breitling and Gerber 2000) and a homeodomain in protozoans. In addition to the original arguments, this scenario seems plausible also because no Pax genes have been found in protists and the best-supported Pax paired domain phylogeny is obtained with transposase as out-group. The diploblast Pax gene tree (fig. 3A and B) is in accordance with a proposed basal position of Placozoa (for overview and references see Syed and Schierwater 2002). If Porifera were basal, however, the Pax evolution model would require a slight modification. Here, the last common ancestor of Placozoa and Porifera would be assumed to have harbored a Trichoplax-type PaxB gene, whose structure remained unchanged in Placozoa but experienced a partial loss of the homeodomain in the lineage leading to Porifera. Because it is known that Placozoa are not secondarily reduced Cnidaria (Ender and Schierwater 2003) and because their bauplan cannot easily be derived from a sponge bauplan (Syed and Schierwater 2002), one has to argue that Pax genes of the A class, and—particularly interesting—also of the B and C class, predated the invention of nerve and sensory cells.
With respect to the evolution of PaxD-like genes additional research is needed to decide whether PaxD branched off even earlier (as shown in our evolution model) or if the PaxD paired domain originated in the next common ancestor of Cnidaria and triploblasts. In the first scenario a PaxD-type gene either got lost or escaped surveys in Placozoa and Porifera. In the second scenario insufficient taxon sampling or nonoptimal out-group choice may have hindered phylogenetic resolution in the analysis. Although it seems unlikely that a PaxD-type gene escaped our PCR screen, we cannot rule out, however, that Placozoans posses more than one Pax gene. To decide between the alternatives more data will be needed, which will likely come from ongoing whole-genome sequencing efforts in Placozoa, Porifera, and Cnidaria.
Functional information from triploblast Pax genes may also add to our understanding of early duplication events in diploblastic animals. It was previously assumed that PaxB is a precursor of Pax2/5/8, whereas PaxC could be a precursor of Pax6 genes in triploblasts. Plaza et al. (2003) recently demonstrated that DNA-binding characteristics of cnidarian PaxB and PaxC proteins display no simple relationship to Pax2/5/8 and Pax6 genes. The authors showed that A. millepora PaxB and PaxC proteins can both bind to eyeless (ey) targets in vivo and in vitro, which casts doubt on the postulated direct relationship between cnidarian Pax genes and the bilaterian Pax6 and Pax2/5/8 classes. Given that our analysis suggests that Cnidarian PaxB and PaxC (and also PaxA) genes are derived from a gene with similar organization to the placozoan PaxB-like gene, one could speculate that the ancestral PaxABC gene unites functional features which were retained in cnidarian PaxB and PaxC as well as in triploblast Pax2/5/8 and Pax6 genes. This hypothesis is indeed supported by Kozmik et al. (2003) who showed that T. cystophora PaxB (the only Pax gene in cubomedusa found so far) contains a Pax2/5/8-type paired domain and octapeptide but a Pax6-type homeodomain. The Tripedalia PaxB gene is expressed in larval stages, in the retina, lens, and statocyst. According to functional properties, that is, binding specificity, the ability to rescue spa (a Drosphila Pax2 mutant) and to induce ectopic eyes in Drosophila, the authors suggest that the ancestor of the cubozoan PaxB-like protein was the primordial Pax protein in eye evolution and that Pax6-like genes evolved in triploblasts after separation from Cnidaria. Trichoplax PaxB meets these expectations for an ancestral Proto-PaxABC gene. Most interestingly, Trichoplax does not possess any kind of sensory organs or nerve cells. Expression of TriPaxB in small irregular cell patches along the outer edge of the animal possibly relates to undifferentiated cells and is spatially overlapping with the Trox2 expression domain (the only Hox/ParaHox gene found in Placozoa; see fig. 6G and Jakob et al. 2004). Quite noteworthy, TriPaxB is not expressed in differentiated fiber cells, which represent putative proto-neural/muscular cells and are located between the upper and lower epithelium throughout the center region of the body (fig. 6C and D). TriPaxB could, however, function in cell determination of fiber cells from undifferentiated and multipotent precursor cells (cf. Jakob et al. 2004). We propose that TriPaxB and Trox2 both demark a particular zone of cell proliferation and differentiation.
We further propose that PaxB was co-opted in the last common ancestor of cnidarians and triploblasts for sensory organ and nerve cell development and that two rounds of gene duplication (B C and C A) followed by partial losses of homeodomain and/or octapeptide sequences led to the current Pax gene assembly in cnidarians (fig. 4). In the triploblast lineage additional duplication-deletion events have taken place and among others resulted in the functional split of protein function of Pax2/5/8 and Pax6 genes (fig. 4). Because TriPaxB is basal to all other known PaxB genes (and also to PaxA and PaxC genes), it is basal also to Pax2/5/8 and Pax6 genes (Sun et al. 1997; fig. 4). Our data suggest that a PaxB similar gene (harboring functional features of both Pax2/5/8 and Pax6 genes) was the original gene involved in sensory organ development and evolution. A functional split into Pax2/5/8 (mechanosensory) and Pax6 (eye/light sense) likely occurred in the last common ancestor of diploblasts and triploblasts. While the placozoan TriPaxB gene most likely predates the origin of nerve and sensory cells, its ancestral developmental function needs to be investigated in more detail.
Acknowledgements
We acknowledge very helpful comments from two anonymous reviewers and support from the Deutsche Forschungsgemeinschaft (DFG Schi 277/10) and Human Frontier Science Program (HFSP RGP0221/2001-M).
References
Balczarek, K. A., Z. C. Lai, and S. Kumar. 1997. Evolution of functional diversification of the paired box (Pax) DNA-binding domains. Mol. Biol. Evol. 14:829–842.
Bone, Q., and K. P. Ryan. 1978. Cupular sense organs in ciona (Tunicata: Ascidiacae). J. Zool. Lond. 186:417–429.
Breitling, R., and J. K. Gerber. 2000. Origin of the paired domain. Dev. Genes Evol. 210:644–650.
Callaerts, P., A. M. Munoz-Marmol, S. Glardon, E. Castillo, H. Sun, W. H. Li, W. J. Gehring, and E. Salo. 1999. Isolation and expression of a Pax-6 gene in the regenerating and intact Planarian Dugesia(G)tigrina. Proc. Natl. Acad. Sci. USA 96:558–563.
Catmull, J., D. C. Hayward, N. E. McIntyre, J. S. Reece-Hoyes, R. Mastro, P. Callaerts, E. E. Ball, and D. J. Miller. 1998. Pax-6 origins—implications from the structure of two coral pax genes. Dev. Genes Evol. 208:352–356.
Czerny, T., G. Halder, U. Kloter, A. Souabni, W. J. Gehring, and M. Busslinger. 1999. Twin of eyeless, a second Pax-6 gene of Drosophila, acts upstream of eyeless in the control of eye development. Mol. Cell 3:297–307.
Ender, A., and B. Schierwater. 2003. Placozoa are not derived cnidarians: evidence from molecular morphology. Mol. Biol. Evol. 20:130–134.
Felsenstein, J. 2003. Inferring Phylogenies. Sinauer Associates, Inc., Sunderland, Mass.
Frigerio, G., M. Burri, D. Bopp, S. Baumgartner, and M. Noll. 1986. Structure of the segmentation gene paired and the Drosophila PRD gene set as part of a gene network. Cell 47:735–746.
Fu, W., H. Duan, E. Frei, and M. Noll. 1998. Shaven and sparkling are mutations in separate enhancers of the Drosophila Pax2 homolog. Development 125:2943–2950.
Fu, W., and M. Noll. 1997. The Pax2 homolog sparkling is required for development of cone and pigment cells in the Drosophila eye. Genes Dev. 11:2066–2078.
Galliot, B., and D. Miller. 2000. Origin of anterior patterning. How old is our head? Trends Genet. 16:1–5.
Groger, H., P. Callaerts, W. J. Gehring, and V. Schmid. 2000. Characterization and expression analysis of an ancestor-type Pax gene in the hydrozoan jellyfish Podocoryne carnea. Mech. Dev. 94:157–169.
Hadrys, T., V. Prince, M. Hunter, R. Baker, and S. Rinkwitz. 2004. Comparative genomic analysis of vertebrate Hox3 and Hox4 genes. J. Exp. Zool. B Mol. Dev. Evol. 302:147–164.
Halder, G., P. Callaerts, and W. J. Gehring. 1995. Induction of ectopic eyes by targeted expression of the eyeless gene in Drosophila. Science 267:1788–1792.
Holland, L. Z., M. Schubert, Z. Kozmik, and N. D. Holland. 1999. AmphiPax3/7, an amphioxus paired box gene: insights into chordate myogenesis, neurogenesis, and the possible evolutionary precursor of definitive vertebrate neural crest. Evol. Dev. 1:153–165.
Hoshiyama, D., H. Suga, N. Iwabe, M. Koyanagi, N. Nikoh, K. Kuma, F. Matsuda, T. Honjo, and T. Miyata. 1998. Sponge Pax cDNA related to Pax-2/5/8 and ancient gene duplications in the Pax family. J. Mol. Evol. 47:640–648.
Huelsenbeck, J. P., and F. Ronquist. 2001. MRBAYES: Bayesian inference of phylogenetic trees. Bioinformatics 17:754–755.
Jakob, W., S. Sagasser, S. Dellaporta, P. Holland, K. Kuhn, and B. Schierwater. 2004. The Trox-2 Hox/ParaHox gene of Trichoplax (Placozoa) marks an epithelial boundary. Dev. Genes Evol. 214:170–175.
Kavaler, J., W. Fu, H. Duan, M. Noll, and J. W. Posakony. 1999. An essential role for the Drosophila Pax2 homolog in the differentiation of adult sensory organs. Development 126:2261–2272.
Kozmik, Z., M. Daube, E. Frei, B. Norman, L. Kos, L. J. Dishaw, M. Noll, and J. Piatigorsky. 2003. Role of Pax genes in eye evolution: a cnidarian PaxB gene uniting Pax2 and Pax6 functions. Dev. Cell 5:773–785.
Kozmik, Z., N. D. Holland, A. Kalousova, J. Paces, M. Schubert, and L. Z. Holland. 1999. Characterization of an amphioxus paired box gene, AmphiPax2/5/8: developmental expression patterns in optic support cells, nephridium, thyroid-like structures and pharyngeal gill slits, but not in the midbrain-hindbrain boundary region. Development 126:1295–1304.
Miller, D. J., D. C. Hayward, J. S. Reece-Hoyes, I. Scholten, J. Catmull, W. J. Gehring, P. Callaerts, J. E. Larsen, and E. E. Ball. 2000. Pax gene diversity in the basal cnidarian Acropora millepora (Cnidaria, Anthozoa): implications for the evolution of the Pax gene family. Proc. Natl. Acad. Sci. USA 97:4475–4480.
O'Brien, E. K., and B. M. Degnan. 2003. Expression of Pax258 in the gastropod statocyst: insights into the antiquity of metazoan geosensory organs. Evol. Dev. 5:572–578.
Plaza, S., D. M. De Jong, W. J. Gehring, and D. J. Miller. 2003. DNA-binding characteristics of cnidarian Pax-C and Pax-B proteins in vivo and in vitro: no simple relationship with the Pax-6 and Pax-2/5/8 classes. J. Exp. Zool. B Mol. Dev. Evol. 299:26–35.
Rinkwitz-Brandt, S., H. H. Arnold, and E. Bober. 1996. Regionalized expression of Nkx5-1, Nkx5-2, Pax2 and sek genes during mouse inner ear development. Hear Res. 99:129–138.
Shimodaira, H., and M. Hasegawa. 1999. Multiple comparisons of log-likelihoods with applications to phylogenetic inference. Mol. Biol. Evol. 16:1114–1116.
———. 2001. CONSEL: for assessing the confidence of phylogenetic tree selection. Bioinformatics 17:1246–1247.
Sun, H., A. Rodin, Y. Zhou, D. P. Dickinson, D. E. Harper, D. Hewett-Emmett, and W. H. Li. 1997. Evolution of paired domains: isolation and sequencing of jellyfish and hydra Pax genes related to Pax-5 and Pax-6. Proc. Natl. Acad. Sci. USA 94:5156–5161.
Swofford, D. L. 2003. PAUP*: phylogenetic analysis using parsimony (*and other methods). Version 4. Sinauer Associates Inc., Sunderland, Mass.
Syed, T., and B. Schierwater 2002. Trichoplax adhaerens: discovered as a missing link, forgotten as a hydrozoan, re-discovered as a key to metazoan evolution. Vie Milieu 52:177–187.
Torres, M., E. Gomez-Pardo, and P. Gruss. 1996. Pax2 contributes to inner ear patterning and optic nerve trajectory. Development 122:3381–3391.
Wada, H., H. Saiga, N. Satoh, and P. W. Holland. 1998. Tripartite organization of the ancestral chordate brain and the antiquity of placodes: insights from ascidian Pax-2/5/8, Hox and Otx genes. Development 125:1113–1122.
Walther, C., J. L. Guenet, D. Simon, U. Deutsch, B. Jostes, M. D. Goulding, D. Plachov, R. Balling, and P. Gruss. 1991. Pax: a murine multigene family of paired box-containing genes. Genomics 11:424–434.(Thorsten Hadrys*,,,1, Rob)