Introduction

The Runt box (Runx) genes encode a family of heteromeric transcription factors, which are characterized by the highly conserved DNA- and protein-binding Runt domain of around 130 amino acids (Kagoshima et al. 1993). Runx genes have been discovered in mammals, nematodes, lancelet fish, and sea urchins (Bae and Lee 2000; Robertson et al. 2002; Stricker et al. 2003). Studies in Drosophila and mouse revealed that Runx transcription factors play important roles in metazoan development (for review, see Coffman 2003). Genetic depletion of the primordial Runx gene runt (run) in Drosophila leads to periodic deletions of larval segments (Gergen and Butler 1988; Gergen and Wieschaus 1985). In addition, Drosophila run plays roles in sex determination and neurogenesis (Duffy et al. 1991; Torres and Sanchez 1992). A second Drosophila Runx gene, lozenge (lz), is required for patterning in the developing antenna and the eye as well as in hematopoiesis for the differentiation of crystal cells (Bataille et al. 2005; Daga et al. 1996; Gupta and Rodrigues 1995).

The Drosophila genome project revealed two additional Runx genes, which are genetically linked with run and lz on the X chromosome (Rennert et al. 2003): CG42267 and CG34145. These genes have not yet been studied in depth, but a genome-wide RNA interference (RNAi) screen identified an isoform of CG42267 to be involved in the control of cell survival (Boutros et al. 2004).

Orthologs of Drosophila run have been described in the red flour beetle Tribolium castaneum and the honeybee Apis mellifera (Choe et al. 2006; Dearden et al. 2006). In addition, the African malaria mosquito Anopheles gambiae was found to possess at least three Runx gene homologs (Rennert et al. 2003). To clarify the evolutionary conservation of Drosophila Runx genes, we searched for orthologs in the genomes of Aedes aegypti (yellow fever mosquito), T. castaneum, A. mellifera, Nasonia vitripennis (parasitoid wasp), Strongylocentrotus purpuratus (sea urchin), and Nematostella vectensis (sea anemone). Our analysis revealed that all Drosophila Runx paralogs are highly conserved in the insect species. Strikingly, the genomic arrangement of insect Runx genes is preserved as well, despite evolutionary turnover of unrelated genes in the same genomic region. These findings imply that insect Runx genes have been conserved as a gene cluster for more than 300 million years, pointing to underlying regulatory constraints. A model of developmental gene cluster evolution is proposed assuming the inherited, and thus shared, need for regulation by a single ancestral enhancer element.

Materials and methods

Sequence retrieval and ortholog search

The Drosophila melanogaster lz, CG42267, CG34145, and run sequences (accession numbers FBgn0002576, FBgn0259162, FBgn0083981, and FBgn0003300) were retrieved from Flybase and used as queries in searches against the genome sequence databases of yellow fever mosquito (A. aegypti genome database version 1.0), red flour beetle (T. castaneum Georgia GA2 genome database version 1.1), honeybee (A. mellifera DH4 genome database version 4.0), parasitic wasp (N. vitripennis genome assembly 1.1), sea urchin (S. purpuratus genome assembly 2.1), sea anemone (N. vectensis genome assembly 1.0), and mouse (Mus musculus genome database version 37.1) with BLASTP or TBLASTX (Altschul et al., 1997). The Branchiostoma lanceolatum (lancelet fish) Runt sequence was obtained through keyword search in the National Center for Biotechnology Information (NCBI) protein database (accession number AAN08565; Stricker et al. 2003).

Multiple sequence alignment

Multiple protein sequence alignments were generated with CLUSTAL W (Thompson et al. 1994) and inspected by eye. Alignment sites with gaps in the Runx domain were eliminated for tree reconstruction except for sites that included gaps in A. gambiae CG42267 and N. vectensis Runt, in which case gaps were due to missing sequence in the open reading frame prediction rather than evolutionary divergence of the protein sequence.

Gene tree reconstruction

Neighbor joining analysis with JTT amino acid substitution model was run in Phylip 3.66 (Felsenstein 2005; Jones et al. 1992). Protein maximum parsimony tree reconstruction was carried out in MEGA version 4.0 (Saitou and Nei 1987; Tamura et al. 2007). TREE-PUZZLE analysis was run with TREE-PUZZLE version 4.0 (Schmidt et al. 2002). Branch support in maximum parsimony and neighbor joining trees was assessed by nonparametric bootstrap on 100 pseudoreplicates (Felsenstein 1986).

Genomic investigation of linkage

Physical gene positions were determined using the NCBI Entrez Map Viewer of the D. melanogaster, T. castaneum, and A. mellifera genomes (http://www.ncbi.nlm.nih.gov/mapview/). In A. aegypti, because of the fragmented status of the available genome (Waterhouse et al. 2008), physical gene positions were retrieved from supercontigs in the NCBI nucleotide database. Supercontigs NW_001810917 and NW_001810728 were linked based on the updated gene model of the A. aegypti CG42267 ortholog, which extends into both contigs (data in Electronic supplementary material).

Results and discussion

Conservation of insect Runx paralogs

To explore the origin of the four Drosophila Runx genes, we performed protein sequence BLAST searches in the genome databases of A. aegypti, T. castaneum, A. mellifera, N. vitripennis, S. purpuratus, and N. vectensis. Outside the insects, our search discovered a new S. purpuratus Runx gene (Spur_Runt2) in addition to the previously published Spur_Runt1 gene (Rennert et al. 2003). Gene tree analysis provided strong support that the mammalian and sea urchin Runx paralogs dated back to independent gene-duplication events (Fig. 1). Only one Runx homolog was found in the sea anemone and the lancelet fish (Stricker et al. 2003). Consistent with previous conclusions (Rennert et al. 2003), these findings demonstrated that the late ancestor of Metazoa possessed a single Runx gene, which experienced independent duplications in at least vertebrates, echinoderms, and arthropods.

Fig. 1
figure 1

Phylogenetic analysis of insect Runx genes. Neighbor-joining tree based on 142 conserved amino acid sites including the Runx domain. Tree branch length was adjusted by likelihood mapping in TREE-PUZZLE version 4.0 with neighbor-joining tree topology as the input. Left branch support numbers reflect neighbor-joining support, middle branch support numbers reflect maximum parsimony support, and right branch support numbers reflect TREE-PUZZLE support. Bar represents 0.05 amino acid substitutions per site. Aaeg, A. aegypti; Amel, A. mellifera; Blan, B. lanceolatum; Dmel, D. melanogaster; Mmus, M. musculus; Nvec, N. vectensis; Nvit, N. vitripennis; Spur, S. purpuratus; Tcas, T. castaneum

Full size image

Four Runx family member genes were found in all investigated insect species. To clarify the phylogenetic relationships between the insect Runx sequences, we generated a multiple alignment of the conserved sequence regions that encompass the Runx domain (data in Electronic supplementary material). Phylogenetic tree estimation strongly supported four orthology groups, each of which included one of the four Drosophila Runx genes (Fig. 1). The deuterostome and cnidarian homologs rooted the insect Runx orthology groups such that the run and lz groups formed a metacluster that was sister to a second metacluster composed of the CG42267 and CG34145 orthology groups. This topology implied that the four insect Runx orthology groups originated by parallel duplications of two ancestral sister paralogs. However, considering the low support of the branch uniting the run and lz groups metacluster (<47), it is also possible that lz, CG42267, and CG34145 originated through consecutive duplications, in which case run is the oldest paralog. Interestingly, the latter hypothesis is also weakly supported in trees in which the Runx homolog of Caenorhabditis elegans (Runt related family member, rnt-1) is included (data in Electronic supplementary material). The C. elegans rnt-1 sequence, however, has experienced extreme substitution rate acceleration and breaks up the monophyly of the Runx cluster, possibly due to a long branch attraction artifact.

Dynamic intron evolution in insect Runx paralogs

Comparative gene structure analysis confirmed the presence of an approximately 40 amino acid long exon at the C-terminal region of the Runx domain flanked by two highly conserved introns (ancient introns 1 and 2 in Fig. 2) (Rennert et al. 2003). The run orthologs of Drosophila, mosquito, red flour beetle, and wasp, however, lacked ancient intron 1, which is still present in the honeybee. Considering the closer relationship of Coleoptera to Diptera than to Hymenoptera (Savard et al. 2006), it is possible that ancient intron 1 was lost in the ancestor of Coleoptera and Diptera. However, this scenario also implies a second loss in the parasitic wasp Nasonia. Further, we noted that the Apis intron-corresponding ancient intron 1 is very small (123 bp). It is therefore also possible that ancient intron 1was lost before the diversification of the endopterygote insect lineages and that Apis convergently regained a small intron at the same position. Requiring two evolutionary changes, this scenario is equally parsimonious as the first, which requires at least two losses (Coleoptera + Diptera, Nasonia). Data from a wider range of Hymenoptera should resolve this issue.

Fig. 2
figure 2

Intron conservation and divergence in the Runx domain. Boxes indicate exons and connecting lines introns. Exons drawn relative to the scale bar. White boxes represent N-terminal Runx domain with variable N-terminal introns. Black boxes represent the highly conserved exon in the C-terminal Runx domain. Dashed line indicates region of incomplete gene model prediction. Gene and species name abbreviations same as in Fig. 1. See text for details

Full size image

Also, the N-terminal region of the Runx domain is characterized by evidence of dynamic intron gain or loss. Most strikingly, in the Hymenoptera lz, CG42267 and CG34145 gene share an intron at the N-terminal end of the Runx domain (Fig. 2). The dipteran lz and CG42267 genes likewise contain introns in this region. Their exact position, however, is only conserved between orthologs but not paralogs. Moreover, no comparable introns are present in the Tribolium Runx homologs. Considering the most consistently supported Runx family tree (Fig. 1), the dipteran lz and CG42267 specific introns were most likely acquired after the diversification of Diptera from Coleoptera and Hymenoptera. The scattered distribution of the position-conserved hymenopteran introns in lz, CG42267 and CG34145, on the other hand, could have been caused by multiple losses of an ancestral intron in Diptera and Coleoptera. Alternatively, concerted evolution may have caused the spread of an intron that originated during hymenopteran evolution. Yet another possibility is that, assuming that run is the oldest paralog, the acquisition of the position-conserved hymenopteran intron could have occurred in the precursor gene to lz, CG42267, and CG34145. Consistent with this, an early divergence of run is the second strongest supported hypothesis in gene tree reconstruction (data in Electronic supplementary material). However, also this scenario implies multiple intron losses in Coleoptera and Diptera after the diversification of major endopterygote lineages. At this point, the sequence of intron loss and gain in the insect Runx genes can only be tentatively reconstructed. The data are unambiguous, however, in highlighting evolutionary turnover of gene structure in the N-terminal Runx domain region.

Conservation of insect Runx paralog linkage

In Drosophila, all four Runx genes map within a region of 11.4 Mb on the X chromosome (Fig. 3). To explore if the genetic linkage of Drosophila Runx genes was also conserved, we investigated the genomic location of Runx genes in Tribolium, Apis, and Aedes (Honeybee Genome Sequencing Consortium 2006; Tribolium Genome Sequencing Consortium 2008). Remarkably, in both Tribolium and Apis, all Runx genes were located within less than 4 Mb on the same chromosome. The region spanning orthologs of CG42267, CG34145, and run was less than 0.16 Mb long in Drosophila, Tribolium, and Apis. Moreover, the arrangement and orientation of CG42267, CG34145, and run orthologs were identical in all four species. Differences in the genomic organization of the Runx paralogs stemmed only from the more variable position of lz. In Drosophila, lz was located 11.2 Mb distally of CG42267 relative to the centromere. In Tribolium, lz was less strongly linked to the other Runx paralogs being separated from CG42267 by 3.86 Mb. In addition, Tribolium lz was inverted compared to the orientation of Drosophila lz. Aedes also possesses an inversion in lz; however, lz is tightly linked to CG42267 (0.25 Mb in distance). In Apis, lz was likewise tightly linked to the other Runx paralogs but proximal to run, thus implying a relocation from one end of the cluster to the other (Fig. 3). Taken together, these data revealed that the close linkage of run with CG42267 and CG34145 in Drosophila is an ancestral aspect of the genomic organization of insect Runx genes. Second, the comparatively weak linkage of lz to the rest of the clustered Runx genes in Drosophila is likely derived as lz is more closely linked in the other insect genome model species including Aedes. However, lz in general is of more variable position than the rest of the Runx genes, even within different Drosophila species (data not shown).

Fig. 3
figure 3

Conserved linkage of insect Runx genes. The number next to the chromosome name under each species name represents the length of region shown in the figure. Runx loci are indicated by blocks colored consistent with Fig. 1. Open reading frame directions indicated by arrows. Non-Runx genes indicated by black boxes except for Tak1, which is indicated by white box. Scale reflects 20 kb in D. melanogaster, T. castaneum, and A. mellifera but 80 kb in A. aegypti. See text for details

Full size image

Evidence of past gene rearrangements in the genomic environment of the Runx paralog arrays

Considering the evidence for conserved microsynteny in the Runx gene-containing genomic region, we investigated the possibility of conserved linkage of additional protein-coding genes (Fig. 3). In the region between CG42267 and run, we found no evidence for further linkage conservation besides that of the Runx paralogs. Apis was unique in containing no additional gene models in the entire Runx gene region. The region between CG42267 and CG34145 was free of further gene models in all species. In the region between CG34145 and run, Tribolium contained the ortholog of the Drosophila yellow-f2 (XP_969206). This arrangement was unique to Tribolium with yellow-f2 located on different chromosomes than the Runx genes in Aedes and Drosophila. In Drosophila, three different genes are contained in the segment between CG34145 and run (Fig. 3). Two of these genes (accession numbers NP_608404 and NP_608405) were Drosophila orphan genes based on the lack of significant BLAST hits in other insect genome models. The third gene was a member of the Cytochrome P450 family (accession number NP_608403). In Aedes, only single gene model (XP_001657549) was found between the CG34145 and run orthologs. This gene model lacked significant similarity to known genes and therefore may likewise be an orphan gene. A second candidate Aedes orphan gene was present between lz and CG42267 together with the ortholog of Drosophila TGF-beta activated kinase 1 (Tak1) (Fig. 3). In both Drosophila and Aedes, the Tak1 gene was located in the Runx gene cluster, distal to CG42267 (Fig. 3). Tak1 is therefore the sole position-conserved gene in the genomic environment of the clustered Runx genes that is not related to the Runx gene family. This situation, however, occurred only to dipteran genomes in the present sample of species. In the four-way genome comparison of Drosophila, Aedes, Tribolium, and Apis, the conservation of microsynteny applied exclusively to the Runx gene paralogs.

Similar degrees of genetic linkage in the Runx and Hox gene clusters

The finding that the close linkage of insect Runx paralogs remained strongly conserved despite rearrangements of other genes in the same region suggested that the Runx genes evolved as a cluster under the impact of linkage-enforcing constraints. The mammalian Runx paralogs Runx1, Runx2, and Runx3 are located on different chromosomes and therefore cannot serve as reference paradigm for assessing the cluster status of insect Runx genes. The paradigm example of conserved microsynteny of closely related members of a developmental gene family is the Hox gene cluster (reviewed in Garcia-Fernandez 2005). To assess if the conserved linkage of insect Runx genes represented similar features, we compared the genomic evolution of the Runx genes with that of the Hox gene cluster in Drosophila, Aedes, Tribolium, and Apis.

We first asked if the members of the Hox and Runx regions exhibited a similar degree of linkage. Aedes was excluded from this analysis because the Aedes Hox gene cluster has not yet been described in detail. Not counting the more extensively rearranged lz paralog, the average intergenic distances between Runx paralogs are 59,122, 35,657, and 36,293 bp in Drosophila, Tribolium, and Apis, respectively, compared to 47,699, 84,141, and 115,890 bp in the Hox complexes of the same species (Brown et al. 2002; Dearden et al. 2006; Drysdale and Crosby 2005; Lewis et al. 2003). Thus taken together, CG42267, CG34145, and run are similarly closely linked like Hox genes in Diptera and even more closely linked in Coleoptera and Hymenoptera. Next, we noted that, similar to the case of the Runx gene array, the Hox gene cluster of some species (such as Apis and Drosophila) is colonized by genes that are not members of the Hox gene family (Dearden et al. 2006; Negre and Ruiz 2007; Shippy et al. 2008). Also similarly, in no case is linkage of these genes conserved (Shippy et al. 2008). This parallel suggested that, in the case of both the Hox and Runx gene regions, constraints act on preserving linkage specifically between the developmental gene family members in spite of recombination events that lead to rearrangements of unrelated genes.

Relaxed developmental gene cluster conservation in Drosophila

The Hox gene cluster of Drosophila is characterized by a higher number of modifications compared to that of Apis and Tribolium. In Drosophila, the Hox genes are divided into the Antennapedia and Bithorax gene complexes (ANT-C and BX-C), while the Hox genes remained preserved in a single array in Anopheles, Tribolium, and Apis (Brown et al. 2002; Dearden et al. 2006; Drysdale and Crosby 2005; Lewis et al. 2003). Second, five transpositions of protein coding genes other than Hox gene family transcription factors were discovered within the Hox cluster of Drosophila, while no transposition events occurred in the mosquito and Tribolium Hox cluster and presumably only one in Apis (Dearden et al. 2006; Negre and Ruiz 2007; Shippy et al. 2008). In combination, these data revealed a striking parallel in the substantially more dramatic diversification of the Hox and Runx gene clusters in Drosophila compared to Tribolium and Apis.

It has been recently concluded that the higher similarity of the Tribolium Hox cluster organization to that in other metazoans indicates that the Tribolium Hox cluster represents a more ancestral organization than the partially dissociated Drosophila Hox cluster (Shippy et al. 2008). As outgroup data are lacking for the Runx gene cluster, we suggest by argument of analogy that the Runx gene clusters of Apis and Tribolium likewise represent more ancestral states of genomic organization. In summary, considering that the insect Runx and Hox gene clusters are consistently similar in their evolutionary dynamics, we conclude that the insect Runx gene array represents a conserved gene family paralog cluster.

A model for the evolution of developmental gene family clusters: differential coduplication of cis-regulatory elements

Our findings lead to the question of whether the clustering of insect Runx genes is of functional significance and, if so, which functional constraints may be responsible. The nature of functional constraints that lead to the preservation of cluster organization has been most extensively pursued in the case of the Hox gene complex (Kmita and Duboule 2003). While a number of factors have been identified in vertebrates, the forces underlying cluster conservation in invertebrates, and thus the most ancestral type of cluster, is still elusive. Indeed, the very existence of such constraints has been questioned (Negre and Ruiz 2007). Recent analysis of the genome of the sea anemone N. vectensis discovered a significant degree of synteny with vertebrate genome structure (Putnam et al. 2007). As long as the reasons for this surprising degree of linkage conservation remain unclear, these data may indicate that the conservation of synteny could reflect rarity of chromosomal rearrangements. This model has been referred to as “phylogenetic inertia” (Negre and Ruiz 2007). However, the deeply conserved synteny in Nematostella applies to genes maintained on corresponding chromosomal scaffolds despite major rearrangements (macrosynteny) rather than the conservation of precise local gene order (microsynteny; Zdobnov and Bork 2007). Further, the genome sequence comparison between Anopheles and Drosophila uncovered only a very limited fraction of genes in conserved synteny (∼30%; Bolshakov et al. 2002; Zdobnov and Bork 2007; Zdobnov et al. 2002). Even more dramatically, only 7% of protein-coding genes were located in conserved microsynteny groups in the comparison of Drosophila, Anopheles, and Apis (Honeybee Genome Sequencing Consortium 2006). These numbers may be conservative in that annotation gaps or mistakes in the more recently published genome drafts may prevent the identification of microsynteny in a significant number of cases. Indeed, the Runx cluster is not included in previous large-scale studies on genome synteny in insects (Bolshakov et al. 2002; Severson et al. 2004; Zdobnov and Bork 2007; Zdobnov et al. 2002). Nonetheless, the available data show that, while macrosynteny is indeed stable over long periods of time, the preservation of precise local gene order as in the case of the Hox and Runx gene clusters is exceptional and hence more likely to be the result of conserving constraints. Of note, evidence for a similar case of developmental paralog cluster conservation has recently been described for the Wnt1, Wnt6, Wnt9, and Wnt10 paralogs in insects (Bolognesi et al. 2008). This suggests that developmental gene cluster organization is a more widespread phenomenon than currently appreciated.

Based on the Hox gene cluster paradigm, chromatin regulation or global enhancers may lead to the conservation of the Runx gene cluster (Kmita and Duboule 2003). In agreement with the second mechanism, recent studies discovered a significant correlation between microsynteny conservation and the presence of highly conserved noncoding sequence elements in insect genomes (Engstrom et al. 2007). Moreover, these genomic regions are significantly enriched with transcriptional regulators of development and thus form genomic regulatory blocks (Engstrom et al. 2007). Unfortunately, we were unable to assess if this is the case for the Runx gene cluster because the orthology of genes sampled in synteny blocks is not yet documented in an accessible manner (Engstrom et al. 2007; Waterhouse et al. 2008). Nonetheless, long-distance regulatory elements like enhancers are the most likely constraining force responsible for this cluster. Functional and comparative genomic studies of CG42267 and CG34145 in Drosophila and other insect models have the potential to reveal cis-regulatory ties between these two genes and run.

We note that future studies of Hox, Wnt, and Runx gene cluster conservation hold the promise to identify general factors of developmental gene cluster conservation. Importantly, the insect Runx cluster is of more recent origin and lower complexity. It should therefore be easier to understand the evolutionary origin of cluster conservation constraints. As a first step in this direction, we propose a model in which the inheritance of regulatory dependence on an ancestral long-distance cis-regulatory element may enforce cluster preservation (Fig. 4). That is, the primordial protostome Runx gene was regulated by a long-distance enhancer element. Mutational events of a large enough scale to duplicate the coding region and closely linked cis-regulatory information but too small to coduplicate the long-distance enhancer would lead to a situation in which the correct transcriptional regulation of the resulting new duplicates would remain dependent on the same long-distance regulatory element. The continued occurrence of such events can be imagined to lead to a string of tandem duplicated developmental gene paralogs, whose recombinatorial flexibility is limited by the reach of the essential enhancer element.

Fig. 4
figure 4

Developmental gene family cluster conservation by differential coduplication of cis-regulatory elements. Triangle represents putative long-distance enhancer. Ellipses represent proximal cis-regulatory elements upstream of the gene. Boxes represent protein-coding regions. Dashed lines indicate regulatory interaction between the putative shared long-distance enhancer and members of the expanding developmental gene cluster. Ancestral and intermediate Runx genes are indicated by black and grey boxes. Subfunctionalized paralogs in the current Runx cluster are colored consistent with Fig. 3. See text for details

Full size image