MBE Advance Access originally published online on June 12, 2007
Molecular Biology and Evolution 2007 24(9):1912-1925; doi:10.1093/molbev/msm120
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Research Articles |
Evolution of the vertebrate twist family and synfunctionalization: a mechanism for differential gene loss through merging of expression domains
Department of Virology and Developmental Genetics, Faculty of Health Sciences, Ben Gurion University of the Negev, Beer Sheva, 84105, Israel
E-mail: gitelman{at}bgu.ac.il.
 |
Abstract |
|---|
 TOP Abstract IntroductionMaterials and MethodsResults and DiscussionNote Added in ProofSupplementary MaterialAcknowledgementsReferences |
|---|
Twist genes are essential for embryonic development and are conserved from jellyfish to human. To study the vertebrate twist family and its evolution, the entire complement of twist genes was obtained for 9 representative species. Phylogenetic analysis showed that a single protochordate twist gene was duplicated at least twice before the teleost-tetrapod split to give rise to 3 ancestral genes, which were further duplicated or deleted, resulting in fluctuating number of twist paralogs in different vertebrate lineages. To find whether changes in gene copy number were associated with changes in gene function, embryonic expression patterns of twist orthologs were evaluated against the number of twist paralogs in different species. The results showed evidence for both neo- and subfunctionalization, and, in addition, for loss of an ancestral regulatory gene. For example, in Xenopus, twist2 was lost, but the twist1 paralog acquired, and therefore preserved, twist2 function. A general model is proposed to explain the data. In this process, termed synfunctionalization, one paralog acquires the expression domain(s) of another. The merging may lead to function shuffle. Alternatively, it may leave one paralog redundant and thus subject to deletion—while its function is retained by the surviving paralog(s). Synfunctionalization is a mechanism that, together with neo- and subfunctionalization, may work to establish equilibrium in the number of genes that regulate developmental processes; it may regulate the complexity of regulatory regions as well as gene copy number and therefore may play a role in evolution of gene function and the structure of genome.
Key Words: twist gene expression • synfunctionalization • subfunctionalization • neofunctionalization • gene duplication and loss • function shuffle
|
Introduction |
|---|
|
TOPAbstractIntroductionMaterials and MethodsResults and DiscussionNote Added in ProofSupplementary MaterialAcknowledgementsReferences |
|---|
Genomes are highly dynamic. Whole genome duplications (e.g., Postlethwait et al. 1998
; Meyer and Schartl 1999; Jaillon et al. 2004) as well as numerous small-scale events have been described, and accumulation and diversification of duplicate genes can be very rapid (e.g., Bailey et al. 2002; Hao and Nei 2004; Redon et al. 2006). Yet the significance and causal relationship between changes in genome complexity and the evolution of novel form and function are not really clear. The idea that duplicate genetic material may be evolutionary fodder (Lewis 1951; Ohno 1970) is intuitively appealing and is quite widely accepted (e.g., Brooke, Garcia-Fernandez, and Holland 1998; Greer et al. 2000; Hoegg et al. 2004). Retention of duplicates may promote acquisition of new functions, increased evolvability or compensation by paralogs of mutant phenotypes (e.g., Ohta 1989; Nowak et al. 1997; Gu et al. 2003).
However, questions remain. For example, it has been suggested that duplication of the teleost genome is responsible for the high diversity of this lineage. Yet whole-genome duplication is not common in tetrapods, which are at least as diverse and varied in form as the teleosts. Also, there is evidence that duplications occur after the development of evolutionary novelty (Hughes et al. 2000) or are not correlated at all with eras of diversification (Suga et al. 1999; Miyata and Suga 2001). More recently Lynch and Conery (2003) suggested that large population sizes, as in most prokaryotes, is a barrier for retention of duplicates. This would altogether uncouple the processes of gene duplication and the resulting increase in genome complexity from increased organism complexity. The duplication-divergence-complementation (DDC) model (Force et al. 1999; Lynch and Force 2000) is attractive because it shows that acquisition of a de novo function is not necessary for maintenance of both duplicates. Instead, complementary deletion of regulatory elements, or subfunctionalization, leaves each copy incapable of assuming the full set of functions of the parent gene and thus is maintained by selection. Numerous studies of paralog-specific expression patterns provide evidence in favor of this process (e.g., Neidert et al. 2001; Dorus et al. 2003; Farber et al. 2003; Santini et al. 2003).
There has been much less discussion of the mechanisms and conditions that allow loss of ancestral genes, i.e., ancient genes with conserved roles (Olson 1999; Wagner 2005), although such loss is well documented in many metazoan lineages (Aravind et al. 2000; Hughes and Friedman 2004; Miller et al. 2005; Nam and Nei 2005). Such genes may disappear in the process of deletion or remodeling of an entire developmental pathway or organ (e.g., Mouchel-Vielh et al. 1998). However, as yet no mechanism equivalent to the DDC model has been described for gene loss, a mechanism that would permit deletion of an ancestral gene without change in developmental program or fitness of an organism.
One approach to understanding genome plasticity is to examine in detail gene families whose members have been well conserved, but whose copy numbers and functions have diverged. Twist genes fit these criteria because they form a small conserved family of basic Helix Loop Helix (bHLH) transcription factors present throughout the metazoa. In Drosophila, the single twist gene plays an essential role in gastrulation, mesoderm differentiation, and in muscle development (Thisse et al. 1987; Baylies and Bate 1996). In mammals, the two paralogs, twist1 and twist2, are first active at later embryonic stages—during the development of the head neural crest (Wolf et al. 1991; Chen and Behringer 1995; Fuchtbauer 1995; Li et al. 1995; Gitelman 1997; O'Rourke and Tam 2002), and neither is important for muscle formation (Chen and Behringer 1995; Gitelman 1997; Sosic et al. 2003).
In this work the complete sets of twist genes in representative chordates were determined and their phylogeny resolved. The data showed that multiple duplications and deletions of twist genes occurred during vertebrate evolution. Embryonic expression patterns, an indicator of function, were then compiled, compared and evaluated against the copy number of twist genes. This led to the formulation of a general mechanism, termed "synfunctionalization" that allows loss of ancestral genes—genes maintained over evolutionary time-scales by positive selection pressure—without change in developmental program.
|
Materials and Methods |
|---|
|
TOPAbstractIntroductionMaterials and MethodsResults and DiscussionNote Added in ProofSupplementary MaterialAcknowledgementsReferences |
|---|
New Twist Genes—Cloning, Sequencing and Data Mining
A low stringency screen of a 48-hour embryonic expression library with zebrafish twist2 bHLH region produced a partial zebrafish twist1a cDNA. Its genomic sequence was found in the pilot library BAC clone zKp106G4. The regions of both the genomic BAC and cDNA containing plasmids were sequenced by standard dye-terminator dideoxy sequencing. Zebrafish twist2 cDNA was provided by Drs Yan and Postlethwait. For sequence analysis and for RNA in situ hybridizations, it was used along with full-length cDNAs for twist1b, twist1a and twist3 cDNAs (Open Biosystems, Inc., Huntsville, AL). Twist genes from various species were identified by BLAST searches (Altschul et al. 1990
) of deduced protein sequences from accessible databases, with the conserved bHLH-containing carboxy region. Sequence data—whether identified as cDNAs, genomic DNA contigs, trace sequences, ESTs, or my own sequencing—were collected, compared and assembled using AssemblyLIGN (Accelrys, San Diego).
Alignment
Deduced twist amino acid sequences were imported into MacClade v4.08 (Maddison and Maddison 2002) and aligned using ClustalX (Jeanmougin et al. 1998) and the standard Gonnet matrix series. In the amino half, since there was no structural information, such as crystallography data, identified functional domains, or predicted secondary structure, the alignment process was based strictly on amino acid sequence. Multiple alignments of the amino halves were made at different gap penalties. Gap opening penalties of 30, 25, 20, 15, 12.5, 10, 7, 5, 5 and gap extension penalties of 1% of these were set and the number of gapped positions, the number of gaps, and the sum of the character alignment scores were calculated. Minor adjustments were made in MacClade manually, typically by sliding one or a few amino acids from one side of a gap to another.
Phylogenetic Analysis
Phylogenetic analyses were done using the PAUP v4.0b10 (Swofford 2003), MrBayes (v3.1.1) (Ronquist and Huelsenbeck 2003), and PHYLIP v3.65 (Felsenstein 2004b) software packages. By default, most of the software routines treat indels as missing data. There is no missing sequence data, but the addition of gap states in the protein alignment to account for indels (fig. 1) represented both a source of unnecessary noise and potential loss of information. The data were therefore refined as follows. Amino-acid positions (characters) were classified as either No-gap (no gaps in any taxa), Gap-run (more than 4 contiguous gap states in a majority of taxa), or Gap-edge (gap state in a minority of taxa at the ends of gap runs or gaps located within conserved blocks) (fig. 1, closed circles (No-gap), open circles (Gap-edge) and open bars (Gap-runs). The information that would be lost by the indel treatment was partially re-encoded by constructing 2 additional character matrices, a 28-character binary matrix representing either the presence or absence of a gap in each taxon at the Gap-edge positions and a 6-character matrix representing the state of each taxon in the 6 Gap-runs (supplementary data). The No-Gap and Gap-edge characters were always included in the analyses and the Gap-run characters always excluded. The 2 additional character matrices were included in the analysis where the software permitted.
|
With PAUP, phylogenetic trees were constructed using the neighbor joining (NJ) (Saitou and Nei 1987
), branch and bound, and the PAUP heuristic methods. For the heuristic methods, both distance and parsimony were set as the optimization criterion. For the other methods, distance was set as the optimization criterion. The mPAM (Xu and Miranker 2004) substitution matrix was expanded to include transitions between the different states of the gap-edge and gap-run matrices, which permitted inclusion of the additional character matrices. Due to a lack of sufficient computational power, the stickleback, tetraodon, and sea urchin taxa were excluded from the branch and bound analysis. An estimate of the robustness of the NJ tree was obtained using a thousand random bootstrap reaquisitions of character data. A maximum likelihood analysis was done using the PROML routine of PHYLIP, without the additional character matrices. A hidden, 4 category, Markov model (1 invariant) and the JTT amino acid substitution frequency matrix (Jones et al. 1992) were used. Random sequence addition and global rearrangements were done for 10 replicates. A Baysian analysis was done using MrBayes. The stickleback, tetraodon, and sea urchin taxa were excluded due to a lack of sufficient computational power. The provision of a multipartition analysis permitted inclusion of the additional character matrices. Amino-acid substitution modeling was the same as for the ML method. Other parameters and running conditions were as recommended by the software developers. Problematic phylogenetic relationships were further resolved using synteny. The 10 genes most closely linked to chosen zebrafish twist homologs were identified. In the other taxa, the chromosomal locations of the orthologs to these twist-linked genes were tabulated and compared to the locations of the twist orthologs.
Wholemount RNA in situ Analysis
Zebrafish embryos were collected and raised at 28.5°C and staged according to Kimmel et al. (1995). Probes for RNA in situ analysis were generated from plasmids each carrying one of the full-length zebrafish twist cDNAs, twist1a, twist1b, twist2, twist3, or short, unique regions from these cDNAs. The plasmids were linearized and dig-labeled transcripts were produced using the Roche (Basel, Switzerland) products and protocols. The transcripts were then subjected to alkaline hydrolysis to obtain probes of about 100 nt in length. A standard zebrafish RNA in situ procedure was followed (Thisse et al. 1993).
|
Results and Discussion |
|---|
|
TOPAbstractIntroductionMaterials and MethodsResults and DiscussionNote Added in ProofSupplementary MaterialAcknowledgementsReferences |
|---|
Twist Family Values
Altogether 33 twist genes, 9 of which had been studied previously and 24 new ones, were identified in 12 representative chordate species by cloning, sequencing, and database searches. Their amino acid sequences are shown in figure 1. Species were chosen both because they represented major vertebrate groups and because their sequencing projects were advanced enough to yield complete gene sets. Twist genes were searched for in mammals (mouse and human), birds (chick, Gallus gallus), amphibia (X. tropicalis and X. laevis), ray-finned fish (zebrafish, Danio rerio; medaka, Oryzias latipes; fugu, Fugu rubripes; stickleback, Gasterosteus aculeatus; pufferfish, Tetraodon nigroviridis), protochordates (Amphioxus, C. intestinalis and C. savigny), and a non-chordate deuterostome, the sea urchin, Strongylocentrotus purpuratus. Twist sequences from other species were not included in the analysis because they were either uninformative (e.g., rat, pig) or too incomplete. The genes were chosen for analysis if their bHLH domain (fig. 1) had over 89% amino acid identity with other twist genes, since this is the level of conservation across most metazoans as far removed as jellyfish and human, and if they contained the highly conserved twist-specific carboxy sequence. This definition was important because it clearly distinguished bona fide twist sequences from related, but non-twist proteins within the large group of bHLH-typeA transcription factors, e.g., N-Twist, Atonal, Ptf1a.
After extensive searches, the number of mouse and human genes remained at 2 each and Amphioxus at 1, the number in frogs went up to 2 and in chicks to 3; 5 twist paralogs were found in medaka and stickleback, 4 in zebrafish and pufferfish, and 3 in fugu. In the sea urchin, only 1 twist gene was identified. In Ciona intestinalis and Ciona savigny, two twist-like genes (45–51% bHLH identity) were picked up in the searches, but no true twist genes could be detected. Regardless of whether they disappeared through deletion or through rapid sequence divergence, the data suggest, as did Hughes and Friedman (2005) when working with a much larger data set, that the urochordates may not be a simple model of chordate development (Passamaneck and Di Gregorio 2005), but may be as divergent from a taxon-wide norm for some developmental processes as the abdomenless barnacles are from canonical arthropod development. The genomes of all these species, except Amphioxus and X. laevis, have been sequenced to at least 6.5X coverage; therefore the gene numbers are probably complete. In the reptiles, one twist gene was identified in WGS trace database of the lizard, Anolis carolinensis; it was kept for analysis as it represented the only twist gene sequenced so far from this major tetrapod branch.
The nomenclature for twist family members (fig. 1, legend) was based on orthology to the mammalian twist1 and twist2 genes as determined by the phylogenetic analyses (below) and according to the vertebrate gene naming guidelines. For example, of the two twist homologs in the frog X. tropicalis, one was orthologous to the mammalian twist1, while the other belonged to a third clade, not found in mammals, and was therefore named twist3.
DNA-cDNA alignments of 8 twist genes from across the vertebrate spectrum showed a conserved genomic organization (fig. 2). In all genes, the first exon contained the entire coding region flanked by a 5'UTR and a short, usually 35 bp long, fragment of the 3'UTR; the second exon contained the remainder of the 3'UTR (fig. 2A). Conservation of boundaries between exons and introns that are located within coding sequences has been well described (Rogozin et. al. 2006). In the case of the twist genes, the conserved exon-intron boundary is within the non-coding 3'UTR (fig. 2A), which is a relatively uncommon position for introns (Hong et al. 2006). Most twist genes have 2 exons, with the exception of human twist2, which has 3 second exons; these are however alternatively used, as shown by analysis of 16 transcripts—such that there is only 1 second exon per transcript. Additional diversity in twist transcripts is generated by alternative use of transcriptional stop sites, which are found in a few second exons (fig. 2, arrows).
|
The sizes of the introns are also conserved, but in a clade-specific manner: introns of twist1 orthologs are 400-600 bp long, while twist2 and twist3 introns are 5 to 150 fold longer. None of the twist genes are tandemly arranged or linked to each other; all are trans-paralogs found on different chromosomes (fig. 2B, brackets).
Phylogeny of the Chordate twist Genes
Alignment of twist amino acid sequences in their conserved carboxy region was unambiguous, while the variable amino-half required the addition of gaps (fig. 1). Different sets of alignments were generated using ClustalX by varying the gap and gap extension penalties and calculating the total number of gap characters, the total number of gaps, and the sum of the aligned character scores. Plotting these against the gap penalties (not shown) showed increasing character scores as the gap penalty was reduced from 25 to 10, whereas the score diminished and the number of gap characters increased sharply below 10. The 15 to 12.5 gap penalty, was therefore chosen for generating the final alignment, which revealed 3 relatively conserved blocks in the amino half (fig. 1), 2 of which are "KR" islands, recognizable as a bipartite nuclear localization signal.
With different phylogenetic reconstruction methods applied to both the full length (fig. 3) and the carboxy region alone ("bHLH trees", not shown), all vertebrate twist genes, fell into three major clades (fig. 3, cladesI–III), each of which contained both teleost and tetrapod genes except for one group of fish genes, the 1b branch (fig. 3 [•]). The neighbor joining (NJ) and PAUP heuristic search methods produced very similar phylogenetic trees (e.g., fig. 3A). Their bootstrap numbers (fig. 3B) showed that most of the branching patterns were robust. The maximum likelihood (ML) analysis (fig. 3C) differed from the NJ and parsimony trees in 2 significant ways. The outgroup amphioxus and sea urchin genes were near the root of cladeII, and both together were rooted in cladeI. This tree presented several unlikely features from the perspective of vertebrate phylogeny. Reconciliation between the gene phylogeny and vertebrate phylogeny (Page and Charleston 1997) would require postulating a large number of gene duplication and deletion events. The Baysian analysis produced a consensus tree (fig. 3D) similar to the NJ tree and provided some insight into the ML results. Some of the highest frequency Baysian trees (not shown) were nearly identical to the ML tree except for the placement of the Amphioxus sequence. However, the posterior probabilities for these trees were below 3.4%, suggesting that little confidence should be placed in the ML tree. "bHLH trees", produced from the unambiguous carboxy region alignments produced similar trees.
|
The very different branch lengths showed that different twist paralogs have been evolving at quite different rates, with cladeIII and the fish 1b genes evolving the fastest. This increased rate was not due to the variability of the amino region since in the "bHLH trees" (not shown) the cladeIII branches were just as long. Because the cladeIII branches of all species are long, the increased rates of sequence divergence cannot be a result of relaxed selection pressure that may immediately follow a gene duplication event (e.g., Li and Gojobori 1983
; Ohta 1993), rather the data suggest that cladeIII has been under comparatively relaxed selection pressure throughout its history.
In other gene families, evolutionary rates of paralog evolution correlate with a variety of gene characteristics, including: retention of ancestral function (Gu et al. 2005; Williams and Holland 1998), severity of phenotype (Ashburner et al. 1999; Llimargas and Lawrence 2001), and restriction of expression (Duret and Mouchiroud 2000; Jordan et al. 2004; Zhang and Li 2004). Indeed, in zebrafish, where both the phylogeny (fig. 4) and the expression patterns of all twist genes have now been established (fig. 5, table 1), it is clear that the faster evolving twist3 gene is expressed more narrowly than twist1a, twist1b or twist2. It is tempting to conceptually tie these findings together by suggesting that the divergence rate of a gene depends on the number of its interacting partners. Since interacting proteins have been shown to co-evolve (Jespers et al. 1999; Lagerstrom et al. 2005), their interactions should act to restrain, or impose an "evolutionary brake" on the overall speed of sequence divergence (Dickerson 1971; Zuckerkandl 1976). However, bioinformatic studies draw conflicting conclusions on this point (Fraser et al. 2002; Bloom and Adami 2003; Jordan et al. 2003) suggesting that either the analyses are insufficiently sophisticated or some additional principle is at work.
|
|
|
The position of the fish 1b genes outside of the 3 main twist groups (fig. 3 [•]) implied that they were the only survivors of a fourth vertebrate clade, a possibility that would require at least 2 independent deletions in both the zebrafish and tetrapod lineages. However, the long branch-lengths suggested that this location was an artifact of long-branch attraction (Felsenstein 1978
) and that they are divergent members of one of the other 3 clades. With respect to sequence similarity, placements within cladeIII and cladeI were equally appropriate. But placement in cladeI (fig. 4) stood out as the most parsimonious choice because it reduced the number of gene duplications/deletions required for consistency with vertebrate phylogeny. This was confirmed by synteny analysis. In a majority of cases, the orthologs of the 10 genes most closely linked to twist1b in zebrafish were also neighbors of the twist1b genes in the other fish species. Moreover, most of the gene neighbors of zebrafish twist1a and twist1b were the neighbors of twist1 genes in the tetrapods (mouse and X. tropicalis). With the caveat that the position of the outgroup is tentative, the tree of figure 4 probably reflects the correct relationships among the vertebrate twist genes and indicates that a single ancestral gene was duplicated at least twice prior to the teleost/tetrapod split to produce three ancestral twist genes which were then further duplicated (like twist1 and twist3 in the teleosts), and/or deleted (like twist2 in the frogs).
Differential Embryonic Expression of the twist Genes
The complex pattern of gain and loss of twist genes raised the question of whether and how these changes in gene number correlate with changes in gene function. All Twist proteins are presumed nuclear transcription factors; therefore, provided they are translated, their function is cell-autonomous and is reflected in their expression domains. I therefore determined the embryonic transcript distribution for the 4 zebrafish genes twist1a, twist1b, twist2 and twist3 (fig. 5) and compared them with the known patterns of twist expression in other tetrapods: mammals (mouse), birds (chick), and amphibia (frog) and in the cephalochordate Amphioxus (table 1). RNA in situ hybridization with each full-length cDNA probe gave a specific signal that was the same, but stronger, than short probes from outside the conserved regions. There was no evidence of cross-hybridization and no signal was produced by sense-strand probes.
All 4 zebrafish twist genes were expressed in either migrating cephalic neural crest or its target structures (but in neither trunk nor tail crest) and in several mesodermal lineages. Fig. 5 shows some examples of both overlap and differences in their expression. In the head, different genes marked different crest subpopulations at different times, in different areas and at different levels. At the early, 6–10 somite stages (fig. 5, A–D), twist1a was present at the level of the hindbrain in the dorsal-most neural keel, even in the premigratory crest cells (fig. 5, A, E). Twist1b marked the lateral border of the keel along the fore, mid- and hindbrain (fig. 5, B, F), while twist2 and twist3 were not expressed at all in the head crest at this stage (fig. 5, C, D, G, H). At the 20–24-somite stage, after epithelial-mesenchymal transformation (EMT) of the crest and its immigration, it was twist1b that became highly expressed in the head crest derivatives, e.g., skull bone primordia, the pharyngeal and branchial arches (fig. 5, J, N). At this stage, twist1a expression spread into the mid-and forebrain areas (fig. 5, I), but its level decreased drastically, while twist2 became expressed slightly in the hindbrain area (fig. 5, K, O) and twist3 appeared in the forebrain region (fig. 5, L, P).
In mesoderm-derived tissues, the expression was also dynamic. At the 6–10 somite stage, only twist2 was expressed in the axial mesoderm and the tail bud (fig. 5, C, G), while only twist1b was expressed in the young somites (not shown). At the 20–24-somite stage, only twist1b was found in myotome (fig. 5, R), while both twist1a and twist2 were expressed in sclerotome (fig. 5, I, Q, K, S) that, as it migrated around the notochord and the neural tube, formed the prevertebrae (fig. 5, U, W). At this stage, only twist2 was found in the hypochord (fig. 5, K, S), only twist1b in the nephric duct (fig. 5, J), and only twist3 in the pectoral fin buds (fig. 5, L, P). Both twist1b and twist2 were expressed in the tail bud region, but in complementary patterns (fig. 5, J, K). Among the four twist paralogs, the twist3 pattern was by far the most restricted in both time and space (fig. 5, compare L vs. I, J, K)—this correlated with an increased rate of divergence of its sequence (fig 4).
The phylogenetic analysis showed repeated instances of twist duplications (fig. 4), and the compiled expression data (table 1) were consistent with twist duplicates being retained through both subfunctionalization (the DDC model) and neofunctionalization. For instance, some expression domains of mouse twist1 were partitioned between the zebrafish co-orthologs twist1a (sclerotome) and twist1b (lateral plate mesoderm). Expression domains of twist1b in the myotome and of twist1b and frog twist1 in the nephric duct (table 1) are unique and may reflect neofunctionalization events. However, it is also possible that these are ancestral functions retained in some species and lost in others. Further evidence may come from the inclusion of more species as data become available and analysis of the dN/dS rate ratios may help understand the selective forces acting on these genes. In summary, fluctuations in the number of twist genes are associated with changes in expression patterns that are consistent with either splitting of the functional load or with acquisition of additional gene functions.
Twist Gene Gain, twist Gene Loss, and Synfunctionalization
There was also loss of ancestral twist genes, which raised the question regarding the fate of their function. For example, genes from all 3 twist clades participate in one of the hallmarks of the vertebrate body plan development—migration and differentiation of neural crest. This is a highly conserved process, and all 3 genes are found in both teleosts and tetrapods, indicating their conservation over a long evolutionary time. Yet, cladeII was lost in amphibia and cladeIII was lost in mammals—raising the question of how these ancient genes, that had been apparently essential for a long time, could disappear. Ancestral gene loss has been described in many lineages (Aravind et al. 2000; Roelofs and Van Haastert 2001; Miller et al. 2005; Hughes and Friedman 2004) and implicated as a mechanism of evolutionary change (Olson 1999; Nam and Nei 2005). Certainly such genes may get lost in the process of deletion or remodeling of a developmental pathway or organ. In the blind marsupial mole, the gene for interphotoreceptor RBP has become a non-functional pseudogene (Springer et al. 1997). Abdominal-class homeotic genes were lost in Cirrepedes (barnacles) that have no abdomen (Mouchel-Vielh et al. 1998). Gene loss can also be envisioned in the context of a single developmental process that is mediated by 2 or more parallel pathways, a feature long ago described by Hans Spemann as "double assurance" (Spemann 1938) and more recently encapsulated within the idea of biochemical buffering (Kitami and Nadeau 2002). Shifts in the relative importance of such pathways have been documented (Lohr et al. 2001; Wilson and Edlund 2001) and a complete shift could leave the genes in one pathway without effect and, therefore, at risk of deletion.
The loss of ancestral genes has typically been discussed in the context of morphogenetic or biochemical pathways changes; however, examination of twist phylogeny and expression (table1, fig. 4) suggested a mechanism that can allow a gene to be lost without the loss of its function. For example, in Amphioxus, the single twist gene is expressed in the notochord—probably an ancient function predating the evolution of vertebrates and the 2 duplications that produced the 3 ancestral twist genes (fig. 4). Among all vertebrates examined, the notochord function is carried out by the twist2 orthologs (table 1) except in the frog, where twist2 was lost and twist1 performs this function instead. It is possible that in amphibia, it was twist1 that retained notochord expression—instead of twist2; this, however, would involve a complex pattern of function losses in other vertebrate lineages. There is another, more parsimonious, explanation: following divergence of the 3 twist genes, the ancient notochord function was retained in the cladeII genes, but lost in the lineage leading to cladesI and III (fig. 6A-"x"). Therefore, the subsequent vertebrate radiation left notochord expression in the different groups—fish, amphibia and amniotes—a unique domain of twist2 (fig. 6A). Later, in amphibia twist1 also acquired notochord expression (fig. 6B), and twist2, now a functionally redundant gene, was deleted (fig. 6C). In this scenario, the loss of a paralog (twist2) was a result of acquisition of a function (notochord expression) by a different paralog (twist1). In keeping with current terminology (Force et al. 1999), I propose synfunctionalization as a term to describe this process of merging functions (fig. 7).
|
|
; X2, 
and 
. (A) Gene loss: In the process of synfunctionalization, the X2 paralog acquires the Proposing a new mechanism immediately raises the issues of its likelihood and its relationship to other processes that affect the fates and functions of duplicated genes within gene families. Synfunctionalization is complementary to both neo- and subfunctionalization. Unlike these 2 mechanisms, which sustain the retention of new paralogs, synfunctionalization permits the loss of an ancestral paralog. There are, however, also certain similarities: Like neofunctionalization, synfunctionalization relies on the ability of a paralog to pick up new expression domains. Like subfunctionalization, synfunctionalization reassigns developmental roles among paralogs without changing developmental processes.
Sub- and synfunctionalization have inverse relationships with respect to promoter complexity. Higher promoter complexity of the original gene increases the likelihood that upon duplication, subfunctionalization will occur and both duplicates will be retained (Force et al. 1999); and, following subfunctionalization, the average complexity of the paralog regulatory regions decreases. In contrast, within a gene family, the paralog with the least complex promoter region will be most likely to be lost following synfunctionalization, because it has the greatest chance to have all its functions represented by the other paralog(s), whose regulatory region complexity would increase. Indeed, developmental genes often have complex regulatory regions and various functions in different tissues, at different expression levels during embryonic and postnatal development. I suggest that the multifunctional regulatory regions are a product of both neo- and synfunctionalization.
These 2 processes counter the effect of the high rate of gene duplication and subfunctionalization, which, otherwise, would eventually partition gene functions, one apiece, among a set of paralogs. While neofunctionalization provides novel functions, synfunctionalization contributes to promoter complexity through merging of functions that already exist in the genome. This merging also supports loss of duplicates and therefore an equilibrium may be reached in the genome where the rate of paralog retention by subfunctionalization is equal to the rate of paralog loss by synfunctionalization. Since synfunctionalization can transfer independent functions among paralogs, it may be an important part of the "bricolage" (Duboule and Wilkins 1998), the "evolutionary tinkering" that defines the genes responsible for the evolving form and function.
Whether synfunctionalization is a likely process depends upon the frequency with which genes may pick up new regulatory elements. This rate is unknown, but both empirical data (McGregor et al. 2001; Bazykin and Kondrashov 2006) and theoretical modeling (Berg et al. 2004) suggest that it may be quite rapid; the ability to gain new regulatory elements forms a key aspect of the neofunctionalization model. Indeed, there are many clear examples of function acquisition among multigene family members in different species, (e.g. Pajovic et al. 1994; Wang et al. 1996; Gu et al. 2005).
Synfunctionalization may also explain the observations in at least 4 different vertebrate gene families where 1 function is executed in different species by different orthologs (Nguyen et al. 1998; Scholpp and Brand 2001; Locascio et al. 2002; Prince 2002). So far explanations of this process, called "function shuffling" by (McClintock et al. 2001), have included complex patterns of subfunctionalization events (Prince 2002), or modulation of gene expression by epigenetic processes (Locascio et al. 2002). However, synfunctionalization offers a simpler mechanism (fig. 7B).
Further data in support of synfunctionalization may come from more complete phylogenies and expression patterns of gene families. For instance, the twist3 paralog, arose before the sarcopterygii-actinopterygii split and has been retained in some amniotes, but was lost in mammals. Whether or not its loss could also be due to synfunctionalization will become clear from comparing embryonic expression patterns of all twist orthologs in the amphibia and birds. Since synfunctionalization affects the complexity of regulatory regions, more concrete evidence at the mechanistic level will come from the comparative analysis of cis-acting regulatory elements.
|
Note Added in Proof |
|---|
|
TOPAbstractIntroductionMaterials and MethodsResults and DiscussionNote Added in ProofSupplementary MaterialAcknowledgementsReferences |
|---|
Satoh et al. (2006, Dev Dyn 235:1065-1073) published a protein sequence that is encoded by the X. laevis twist3 gene.
|
Supplementary Material |
|---|
|
TOPAbstractIntroductionMaterials and MethodsResults and DiscussionNote Added in ProofSupplementary MaterialAcknowledgementsReferences |
|---|
Supplementary material is available at Molecular Biology and Evolution online (http://www.mbe.oxfordjournals.org).
ACCESSION NUMBERS: A. carolinensis twist1 - BK006262 [GenBank] ; D. rerio twist1a - EF620930 [GenBank] ; D. rerio twist1b - BK006285 [GenBank] ; D. rerio twist2 - EF620931 [GenBank] ; D. rerio twist3 - BK006286 [GenBank] ; F. rubripes twist1a - BK006273 [GenBank] ; F. rubripes twist1b - BK006274 [GenBank] ; F. rubripes twist2 - BK006275 [GenBank] ; G. aculeatus twist1a - BK006276 [GenBank] ; G. aculeatus twist1b - BK006277 [GenBank] ; G. aculeatus twist2 - BK006278 [GenBank] ; G. aculeatus twist3a - BK006279 [GenBank] ; G. aculeatus twist3b - BK006280 [GenBank] ; G. gallus twist1 - BK006263 [GenBank] ; G. gallus twist2 - BK006264 [GenBank] ; G. gallus twist3 - BK006265 [GenBank] ; O. latipes twist1a - BK006268 [GenBank] ; O. latipes twist1b - BK006269 [GenBank] ; O. latipes twist2 - BK006270 [GenBank] ; O. latipes twist3a - BK006271 [GenBank] ; O. latipes twist3b - BK006272 [GenBank] ; S. purpuratus twist - BK006287 [GenBank] ; T. nigrovirides twist1a - BK006281 [GenBank] ; T. nigrovirides twist1b - BK006282 [GenBank] ; T. nigrovirides twist2 - BK006283 [GenBank] ; T. nigrovirides twist3 - BK006284 [GenBank] ; X. tropicalis twist1 - BK006266 [GenBank] ; X. tropicalis twist3 - BK006267 [GenBank] .
|
Acknowledgements |
|---|
|
TOPAbstractIntroductionMaterials and MethodsResults and DiscussionNote Added in ProofSupplementary MaterialAcknowledgementsReferences |
|---|
I am sincerely grateful to the people who do the genome sequencing projects, maintain the databases and make the information available: the National Center for Biotechnology Information (NCBI) of the National Institutes of Health, Medaka Genome Initiative, Department of Energy Joint Genome Institute, International Fugu Genome Consortium, Wellcome trust Sanger Institute, the European Bioinformatics Institute (Ensembl), Washington University Zebrafish genome resources.
Many thanks to Drs. Yan and Postlethwait for the zebrafish twist (now twist2) and to Drs Koch (Hubrecht Laboratory) and Humphray (Wellcome Trust Sanger Institute) for providing BACzKp106G4 from the Daniokey Pilot BAC library.
This work was supported by a grant from the Israel/USA Bi-national Science Foundation (BSF).
|
Footnotes |
|---|
William Martin, Associate Editor
|
References |
|---|
|
TOPAbstractIntroductionMaterials and MethodsResults and DiscussionNote Added in ProofSupplementary MaterialAcknowledgementsReferences |
|---|
Altschul SF, Gish W, Miller W, Myers EW, Lipman DJ. Basic local alignment search tool. J Mol Biol. (1990) 215:403–410.[CrossRef][Web of Science][Medline]
Aravind L, Watanabe H, Lipman DJ, Koonin EV. Lineage-specific loss and divergence of functionally linked genes in eukaryotes. Proc Natl Acad Sci USA (2000) 97:11319–11324.
Ashburner M, Misra S, Roote J, et al. An exploration of the sequence of a 2.9-Mb region of the genome of Drosophila melanogaster: the Adh region. Genetics (1999) 153:179–219.
Bailey JA, Gu Z, Clark RA, Reinert K, Samonte RV, Schwartz S, Adams MD, Myers EW, Li PW, Eichler EE. Recent segmental duplications in the human genome. Science (2002) 297:1003–1007.
Baylies MK, Bate M. Twist: a myogenic switch in Drosophila. Science (1996) 272:1481–1484.[Abstract]
Bazykin GA, Kondrashov AS. Rate of promoter class turn-over in yeast evolution. BMC Evol Biol. (2006) 6:14.[CrossRef][Medline]
Berg J, Willmann S, Lassig M. Adaptive evolution of transcription factor binding sites. BMC Evol Biol. (2004) 4:42.[CrossRef][Medline]
Bloom JD, Adami C. Apparent dependence of protein evolutionary rate on number of interactions is linked to biases in protein-protein interactions data sets. BMC Evol Biol. (2003) 3:21.[CrossRef][Medline]
Brooke NM, Garcia-Fernandez J, Holland PW. The ParaHox gene cluster is an evolutionary sister of the Hox gene cluster. Nature (1998) 392:920–922.[CrossRef][Medline]
Chen ZF, Behringer RR. Twist is required in head mesenchyme for cranial neural tube morphogenesis. Genes Dev. (1995) 9:686–699.
Dickerson RE. The structures of cytochrome c and the rates of molecular evolution. J Mol Evol. (1971) 1:26–45.[CrossRef][Medline]
Dorus S, Gilbert SL, Forster ML, Barndt RJ, Lahn BT. The CDY-related gene family: coordinated evolution in copy number, expression profile and protein sequence. Hum Mol Genet. (2003) 12:1643–1650.
Duboule D, Wilkins AS. The evolution of "bricolage. Trends Genet. (1998) 14:54–59.[CrossRef][Web of Science][Medline]
Duret L, Mouchiroud D. Determinants of substitution rates in mammalian genes: expression pattern affects selection intensity but not mutation rate. Mol Biol Evol. (2000) 17:68–74.
Farber SA, De Rose RA, Olson ES, Halpern ME. The zebrafish annexin gene family. Genome Res. (2003) 13:1082–1096.
Felsenstein J. Cases in which parsimony or compatibility methods will be positively misleading. System Zool (1978) 27:401–410.[CrossRef]
Felsenstein J. Inferring phylogenies (2004a) New York, NY: Sinauer Associates.
Felsenstein J. PHYLIP (Phylogeny Inference Package). Department of Genomic Sciences (2004b) Seattle, WA: University of Washington.
Force A, Lynch M, Pickett FB, Amores A, Yan YL, Postlethwait J. Preservation of duplicate genes by complementary, degenerative mutations. Genetics (1999) 151:1531–1545.
Fraser HB, Hirsh AE, Steinmetz LM, Scharfe C, Feldman MW. Evolutionary rate in the protein interaction network. Science (2002) 296:750–752.
Fuchtbauer EM. Expression of M-twist during postimplantation development of the mouse. Dev Dyn (1995) 204:316–322.[Web of Science][Medline]
Germanguz I, Lev D, Waisman T, Kim CH, Gitelman I. Four twist genes in zebrafish, four expression patterns. Dev Dyn (2007) 236:2615–2626.[Medline]
Gitelman I. Twist protein in mouse embryogenesis. Dev Biol. (1997) 189:205–214.[CrossRef][Web of Science][Medline]
Greer JM, Puetz J, Thomas KR, Capecchi MR. Maintenance of functional equivalence during paralogous Hox gene evolution. Nature (2000) 403:661–665.[CrossRef][Medline]
Gu X, Zhang Z, Huang W. Rapid evolution of expression and regulatory divergences after yeast gene duplication. Proc Natl Acad Sci USA (2005) 102:707–712.
Gu Z, Steinmetz LM, Gu X, Scharfe C, Davis RW, Li WH. Role of duplicate genes in genetic robustness against null mutations. Nature (2003) 421:63–66.[CrossRef][Medline]
Hao L, Nei M. Genomic organization and evolutionary analysis of Ly49 genes encoding the rodent natural killer cell receptors: rapid evolution by repeated gene duplication. Immunogenetics (2004) 56:343–354.[Web of Science][Medline]
Hoegg S, Brinkmann H, Taylor JS, Meyer A. Phylogenetic timing of the fish-specific genome duplication correlates with the diversification of teleost fish. J Mol Evol. (2004) 59:190–203.[CrossRef][Web of Science][Medline]
Hong X, Scofield DG, Lynch M. Intron size, abundance, and distribution within untranslated regions of genes. Mol Biol Evol. (2006) 23:2392–2404.
Hopwood ND, Pluck A, Gurdon JB. A Xenopus mRNA related to Drosophila twist is expressed in response to induction in the mesoderm and the neural crest. Cell. (1989) 59:893–903.[CrossRef][Web of Science][Medline]
Hughes AL, Friedman R. Loss of ancestral genes in the genomic evolution of Ciona intestinalis. Evol Dev. (2005) 7:196–200.[CrossRef][Web of Science][Medline]
Hughes AL, Friedman R. Shedding genomic ballast: extensive parallel loss of ancestral gene families in animals. J Mol Evol. (2004) 59:827–833.[CrossRef][Web of Science][Medline]
Hughes AL, Green JA, Garbayo JM, Roberts RM. Adaptive diversification within a large family of recently duplicated, placentally expressed genes. Proc Natl Acad Sci USA (2000) 97:3319–3323.
Jaillon O, Aury JM, Brunet F, et al. Genome duplication in the teleost fish Tetraodon nigroviridis reveals the early vertebrate proto-karyotype. Nature (2004) 431:946–957.[CrossRef][Medline]
Jeanmougin F, Thompson JD, Gouy M, Higgins DG, Gibson TJ. Multiple sequence alignment with Clustal X. Trends Biochem Sci. (1998) 23:403–405.[CrossRef][Web of Science][Medline]
Jespers L, Lijnen HR, Vanwetswinkel S, Van Hoef B, Brepoels K, Collen D, De Maeyer M. Guiding a docking mode by phage display: selection of correlated mutations at the staphylokinase-plasmin interface. J Mol Biol. (1999) 290:471–479.[CrossRef][Web of Science][Medline]
Jones DT, Taylor WR, Thornton JM. The rapid generation of mutation data matrices from protein sequences. Comput Appl Biosci (1992) 8:275–282.
Jordan IK, Marino-Ramirez L, Wolf YI, Koonin EV. Conservation and coevolution in the scale-free human gene coexpression network. Mol Biol Evol. (2004) 21:2058–2070.
Jordan IK, Wolf YI, Koonin EV. No simple dependence between protein evolution rate and the number of protein-protein interactions: only the most prolific interactors tend to evolve slowly. BMC Evol Biol. (2003) 3:1.[CrossRef][Medline]
Kimmel CB, Ballard WW, Kimmel SR, Ullmann B, Schilling TF. Stages of embryonic development of the zebrafish. Dev Dyn (1995) 203:253–310.[Web of Science][Medline]
Kitami T, Nadeau JH. Biochemical networking contributes more to genetic buffering in human and mouse metabolic pathways than does gene duplication. Nat Genet. (2002) 32:191–194.[CrossRef][Web of Science][Medline]
Lagerstrom MC, Fredriksson R, Bjarnadottir TK, et al. Origin of the prolactin-releasing hormone (PRLH) receptors: evidence of coevolution between PRLH and a redundant neuropeptide Y receptor during vertebrate evolution. Genomics (2005) 85:688–703.[CrossRef][Web of Science][Medline]
Lewis EB. Pseudoallelism and gene evolution. Cold Spring Harbour Symp Quant Biol. (1951) 16:159–174.
Li L, Cserjesi P, Olson EN. Dermo-1: a novel twist-related bHLH protein expressed in the developing dermis. Dev Biol. (1995) 172:280–292.[CrossRef][Web of Science][Medline]
Li WH, Gojobori T. Rapid evolution of goat and sheep globin genes following gene duplication. Mol Biol Evol. (1983) 1:94–108.[Abstract]
Llimargas M, Lawrence PA. Seven Wnt homologues in Drosophila: a case study of the developing tracheae. Proc Natl Acad Sci USA (2001) 98:14487–14492.
Locascio A, Manzanares M, Blanco MJ, Nieto MA. Modularity and reshuffling of Snail and Slug expression during vertebrate evolution. Proc Natl Acad Sci USA (2002) 99:16841–16846.
Lohr U, Yussa M, Pick L. Drosophila fushi tarazu. a gene on the border of homeotic function. Curr Biol. (2001) 11:1403–1412.[CrossRef][Web of Science][Medline]
Lynch M, Conery JS. The origins of genome complexity. Science (2003) 302:1401–1404.
Lynch M, Force A. The probability of duplicate gene preservation by subfunctionalization. Genetics (2000) 154:459–473.
Maddison DR, Maddison WP. MacClade (2002) Sunderland, MA: Sinauer Associates.
McClintock JM, Carlson R, Mann DM, Prince VE. Consequences of Hox gene duplication in the vertebrates: an investigation of the zebrafish Hox paralogue group 1 genes. Development (2001) 128:2471–2484.
McGregor AP, Shaw PJ, Hancock JM, Bopp D, Hediger M, Wratten NS, Dover GA. Rapid restructuring of bicoid-dependent hunchback promoters within and between Dipteran species: implications for molecular coevolution. Evol Dev. (2001) 3:397–407.[CrossRef][Web of Science][Medline]
Meyer A, Schartl M. Gene and genome duplications in vertebrates: the one-to-four (to-eight in fish) rule and the evolution of novel gene functions. Curr Opin Cell Biol. (1999) 11:699–704.[CrossRef][Web of Science][Medline]
Miller DJ, Ball EE, Technau U. Cnidarians and ancestral genetic complexity in the animal kingdom. Trends Genet. (2005) 21:536–539.[CrossRef][Web of Science][Medline]
Miyata T, Suga H. Divergence pattern of animal gene families and relationship with the Cambrian explosion. Bioessays (2001) 23:1018–1027.[CrossRef][Web of Science][Medline]
Mouchel-Vielh E, Rigolot C, Gibert JM, Deutsch JS. Molecules and the body plan: the Hox genes of Cirripedes (Crustacea). Mol Phylogenet Evol. (1998) 9:382–389.[CrossRef][Web of Science][Medline]
Nam J, Nei M. Evolutionary change of the numbers of homeobox genes in bilateral animals. Mol Biol Evol. (2005) 22:2386–2394.
Neidert AH, Virupannavar V, Hooker GW, Langeland JA. Lamprey Dlx genes and early vertebrate evolution. Proc Natl Acad Sci USA (2001) 98:1665–1670.
Nguyen VH, Schmid B, Trout J, Connors SA, Ekker M, Mullins MC. Ventral and lateral regions of the zebrafish gastrula, including the neural crest progenitors, are established by a bmp2b/swirl pathway of genes. Dev Biol. (1998) 199:93–110.[CrossRef][Web of Science][Medline]
Nowak MA, Boerlijst MC, Cooke J, Smith JM. Evolution of genetic redundancy. Nature (1997) 388:167–171.[CrossRef][Medline]
O'Rourke MP, Tam PP. Twist functions in mouse development. Int J Dev Biol. (2002) 46:401–413.[Web of Science][Medline]
Ohno S. Evolution by gene duplication (1970) Berlin: Springer Verlag.
Ohta T. Pattern of nucleotide substitutions in growth hormone-prolactin gene family: a paradigm for evolution by gene duplication. Genetics (1993) 134:1271–1276.[Abstract]
Ohta T. Role of gene duplication in evolution. Genome (1989) 31:304–310.[Medline]
Olson MV. When less is more: gene loss as an engine of evolutionary change. Am J Hum Genet. (1999) 64:18–23.[CrossRef][Web of Science][Medline]
Page RD, Charleston MA. From gene to organismal phylogeny: reconciled trees and the gene tree/species tree problem. Mol Phylogenet Evol. (1997) 7:231–240.[CrossRef][Web of Science][Medline]
Pajovic S, Jones VE, Prowse KR, Berger FG, Baumann H. Species-specific changes in regulatory elements of mouse haptoglobin genes. J Biol Chem. (1994) 269:2215–2224.
Passamaneck YJ, Di Gregorio A. Ciona intestinalis: chordate development made simple. Dev Dyn (2005) 233:1–19.[CrossRef][Web of Science][Medline]
Postlethwait JH, Yan YL, Gates MA, et al. Vertebrate genome evolution and the zebrafish gene map. Nat Genet. (1998) 18:345–349.[CrossRef][Web of Science][Medline]
Redon R, Ishikawa S, Fitch KR, et al. Global variation in copy number in the human genome. Nature (2006) 444:444–454.[CrossRef][Medline]
Roelofs J, Van Haastert PJ. Genes lost during evolution. Nature (2001) 411:1013–1014.[Medline]
Rogozin I, Sverdlov AV, Babenko VN, Koonin EV. Analysis of evolution of exon-intron structure of eukaryotic genes. Brief Bioinform (2005) 6:118–134.
Ronquist F, Huelsenbeck JP. MrBayes 3: Bayesian phylogenetic inference under mixed models. Bioinformatics (2003) 19:1572–1574.
Saitou N, Nei M. The neighbor-joining method: a new method for reconstructing phylogenetic trees. Mol Biol Evol. (1987) 4:406–425.[Abstract]
Santini S, Boore JL, Meyer A. Evolutionary conservation of regulatory elements in vertebrate Hox gene clusters. Genome Res. (2003) 13:1111–1122.
Scaal M, Fuchtbauer EM, Brand-Saberi B. cDermo-1 expression indicates a role in avian skin development. Anat Embryol (Berl) (2001) 203:1–7.[CrossRef][Medline]
Scholpp S, Brand M. Morpholino-induced knockdown of zebrafish engrailed genes eng2 and eng3 reveals redundant and unique functions in midbrain–hindbrain boundary development. Genesis (2001) 30:129–133.[CrossRef][Web of Science][Medline]
Sosic D, Richardson JA, Yu K, Ornitz DM, Olson EN. Twist regulates cytokine gene expression through a negative feedback loop that represses NF-kappaB activity. Cell. (2003) 112:169–180.[CrossRef][Web of Science][Medline]
Spemann H. Embryonic development and induction (1938) New Haven: Yale University Press.
Springer MS, Burk A, Kavanagh JR, Waddell VG, Stanhope MJ. The interphotoreceptor retinoid binding protein gene in therian mammals: implications for higher level relationships and evidence for loss of function in the marsupial mole. Proc Natl Acad Sci USA (1997) 94:13754–13759.
Suga H, Koyanagi M, Hoshiyama D, Ono K, Iwabe N, Kuma K, Miyata T. Extensive gene duplication in the early evolution of animals before the parazoan-eumetazoan split demonstrated by G proteins and protein tyrosine kinases from sponge and hydra. J Mol Evol. (1999) 48:646–653.[CrossRef][Web of Science][Medline]
Swofford DL. PAUP*. In: Phylogenetic Analysis Using Parsimony (*and Other Methods) (2003) Sunderland, MS: Sinauer Associates. Version 4.
Tavares AT, Izpisuja-Belmonte JC, Rodriguez-Leon J. Developmental expression of chick twist and its regulation during limb patterning. Int J Dev Biol. (2001) 45:707–713.[Web of Science][Medline]
Thisse B, el Messal M, Perrin-Schmitt F. The twist gene: isolation of a Drosophila zygotic gene necessary for the establishment of dorsoventral pattern. Nucleic Acids Res. (1987) 15:3439–3453.
Thisse C, Thisse B, Schilling TF, Postlethwait JH. Structure of the zebrafish snail1 gene and its expression in wild-type, spadetail and no tail mutant embryos. Development (1993) 119:1203–1215.[Abstract]
Wagner A. Distributed robustness versus redundancy as causes of mutational robustness. Bioessays (2005) 27:176–188.[CrossRef][Web of Science][Medline]
Wang D, Marsh JL, Ayala FJ. Evolutionary changes in the expression pattern of a developmentally essential gene in three Drosophila species. Proc Natl Acad Sci USA (1996) 93:7103–7107.
Williams NA, Holland PW. Gene and domain duplication in the chordate Otx gene family: insights from amphioxus Otx. Mol Biol Evol. (1998) 15:600–607.[Abstract]
Wilson SI, Edlund T. Neural induction: toward a unifying mechanism. Nat Neurosci (2001) 4(Suppl):1161–1168.[Web of Science][Medline]
Wolf C, Thisse C, Stoetzel C, Thisse B, Gerlinger P, Perrin-Schmitt F. The M-twist gene of Mus is expressed in subsets of mesodermal cells and is closely related to the Xenopus X-twi and the Drosophila twist genes. Dev Biol. (1991) 143:363–373.[CrossRef][Web of Science][Medline]
Xu W, Miranker DP. A metric model of amino acid substitution. Bioinformatics (2004) 20:1214–1221.
Yasui K, Zhang SC, Uemura M, Aizawa S, Ueki T. Expression of a twist-related gene, Bbtwist, during the development of a lancelet species and its relation to cephalochordate anterior structures. Dev Biol. (1998) 195:49–59.[CrossRef][Web of Science][Medline]
Zhang L, Li WH. Mammalian housekeeping genes evolve more slowly than tissue-specific genes. Mol Biol Evol. (2004) 21:236–239.
Zuckerkandl E. Evolutionary processes and evolutionary noise at the molecular level. I. Functional density in proteins. J Mol Evol. (1976) 7:167–183.[CrossRef][Web of Science][Medline]
CiteULike 
Connotea 
Del.icio.us What's this?
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||