Higher phenotypic complexity of vertebrates has been often associated with a higher number of genes produced through whole genome duplications 1. Genome projects and a deluge of sequence data showed that vertebrates often possess more than one copy of a gene or gene clusters 2345 where invertebrates have only one. This observation together with synteny data, led to the formulation of the 2R hypothesis, which proposes two rounds of genome duplication in early vertebrate evolution 6789. An additional duplication event occurred in the lineage of ray-finned fish, the so-called fish-specific genome duplication (FSGD, 3R) 1011121314151617. While the duplicated genes are expected to be redundant in their function immediately following the duplication, their functions often diversify later 1819. Possible scenarios are that one copy evolves a new function (neofunctionalization) or the ancestral functions get subdivided between the paralogs (subfunctionalization) 2021. In most cases, however, one copy is expected to accumulate mutations that lead to a non-functional gene and finally to gene loss 20212223.

Clusters of genes belonging to the same gene family can give important insights into the evolutionary history of a genomic region, both in terms of gene loss events as well as for the evolution of the regulatory sequences surrounding it. The most prominent examples for this type of approach are the Hox gene clusters 2425262728, a family of transcription factors that are not only arranged in uninterrupted clusters on the chromosome but are even expressed during embryogenesis according to their chromosomal order – a phenomenon called colinearity [reviewed in 29. But also other gene clusters have been studied in this regard such as the ParaHox cluster 30(Siegel et al. submitted) and Fox clusters 31. Both belong to other families of transcription factors with multiple cluster copies in vertebrate genomes. Less research in this respect has been performed on non-developmental genes.

We were interested if the patterns of molecular evolution that are found in Hox clusters can also be identified in other gene clusters and if it is also possible to identify conserved non-coding regions in them. KCNA genes are arranged in two uninterrupted clusters of three genes each which are located on chromosomes three and six in mouse and on chromosomes one and twelve in humans 3233. KCNA genes code for the Kv1 family of shaker-related voltage-gated potassium (K⁺) channels, those consist of six transmembrane (TM) segments and, the most important part, the pore loop (P-region), which ensures ion selectivity 3435. The Kv channels are active as tetramers, usually heterotetramers. Sodium (Na⁺) and calcium (Ca2⁺) channels, on the other hand, are monomers that consist of four linked domains, each of which is homologous to a single 6-TM K⁺ channel 36. Studies on the genomic organization of these genes so far have been limited to mammals. Upstream regulatory factors and potential regulatory elements have not been described previously. The number of KCNA genes and their genomic arrangement in ray-finned fish has not been studied before. Here we extend these comparative genomic approaches to other lineages of vertebrates and compare them to the situation in the genome of invertebrates.

We conducted an analysis of complete genome sequences of tetrapods and ray-finned fish KCNA genes and investigated the entire genome content when possible. In an effort to increase the database on basal fish, for which such data were not available prior to this study, we added new data using a PCR approach with universal primers and cloned the PCR products. We also included data from the non-vertebrate chordates Branchiostoma floridae and Ciona intestinalis for a better estimate of the age of this gene family. We constructed gene trees to permit inferences about the timing of the gene duplications and the cluster duplications. Furthermore, we aimed to test the hypothesis that the conservation of the genomic architecture of a gene cluster is linked to the content of conserved elements within the intergenic regions. To this end we investigated the 3-gene-cluster of KCNA genes (Kv1, shaker-related potassium channels) in several species of vertebrates. In tetrapods, two clusters exist (KCNA3-KCNA2-KCNA10, KCNA6-KCNA1-KCNA5), while teleosts were found to have four clusters.

Tetrapods, such as human, chicken and frog, have two three-gene-clusters (3-2-10, 6-1-5) and two additional genes, KCNA4 and KCNA7, which are located elsewhere in the genome (Figure 1). Teleost fish such as pufferfish, medaka, stickleback and zebrafish were found to have four clusters of KCNA genes. According to available results from genome sequencing projects, KCNA5a was lost. For KCNA7, duplicates were found in medaka and two copies of KCNA4 are present in the osteoglossomorph elephantnose fish (Gnathonemus petersi). All of these genes are conserved in their transmembrane domains and in the pore-loop region, but the other parts of these genes are highly variable and impossible to align between different members of this gene family. In the tunicate Ciona intestinalis only one KCNA gene was found while we received several BLAST hits for the amphioxus (Branchiostoma floridae) genome.

A phylogenetic analysis of all KCNA genes suggests a diversification of KCNA genes in the vertebrates and basal position of KCNA7 among them (Figure 2); it is the only vertebrate KCNA gene that contains a single intron while all others are intronless. In invertebrates, such as Ciona intestinalis and Drosophila melanogaster, the KCNA/shaker gene has multiple introns, but the position of the introns is not conserved between invertebrates and vertebrates (data not shown). Genes identified from the Branchiostoma floridae genome, however, consisted of a single exon indicating that the vertebrate KCNA7 gene acquired its intron independently (Figure 3). The three best BLAST hits from Branchiostoma floridae with vertebrate KCNA genes were included in the phylogenetic analysis. All three genes are positioned on different scaffolds and no clusters were found. We obtained additional hits with the Branchiostoma genome, some of which also received KCNA best hits in BLAST searches against the GenBank database. However, their high sequence divergence made reliable estimation of their phylogenetic position relative to the outgroup in the KCNA gene tree impossible. The three Branchiostoma KCNA genes included in the analysis form a monophyletic group basal to the vertebrate genes, indicating a series of independent gene duplication within the amphioxus lineage.

Within the vertebrate part of the tree, KCNA5 and KCNA10 form a monophyletic group, as do KCNA1 and KCNA2 (Figure 2). This finding supports the hypothesis that the two clusters are the result of a complete duplication of the original cluster rather than of independent tandem duplications, which would have phylogenetically grouped neighboring genes on chromosomes in the tree. Only the KCNA3/KCNA6 gene pair does not reflect a pattern of whole genome duplication(s).

Our PCR based approach yielded two KCNA genes from Hydrolagus colliei (spotted ratfish, KCNA2, 5), four genes from Acipenser baerii (sturgeon, KCNA1, 2, 6, 10), six genes from Lepisosteus platyrhynchus (gar, KCNA1, 2, 3, 5, 6, 10), and nine genes from Gnathonemus petersii (elephantnose fish, KCNA1b, 2a?, 3b, 4a, 4b, 5b, 6b, 10a, 10b) [for accession numbers see Additional file 4]. We performed phylogenetic analyses of the ancient duplicates (KCNA3-6, KCNA2-1 and KCNA10-5) and used KCNA4 as outgroup [see Additional files 1, 2, 3]. KCNA4 is not part of the clusters, but phylogenetically closely related without obvious acceleration of evolutionary rates (Figure 2). In this way, we avoid reconstruction artifacts due to a too divergent outgroup as would expected be with the ancestral KCNA7 gene.

Our KCNA3/6 dataset encompassed 42 sequences and included sequences from human, chick and frog (378 amino acid positions) and resulted in Maximum Likelihood and Bayesian inference trees which were congruent for the well supported nodes and showed only minor differences within the not strongly resolved parts of the tree [see Additional file 1]. While the KCNA3 genes were phylogenetically separate from the KCNA6 genes, the resolution within each of these sets of orthologous genes is poor and the relationships, especially among the fish paralogous groups, could not be identified with confidence. The assignment into "a" and "b" paralogs was done based on the position in the clusters when genomic data was available, and the newly determined orthologous KCNA genes for the fish were assigned names accordingly. The evolutionary rates differ clearly between the orthologous groups (KCNA3 vs. KCNA6) as well as between the fish-specific a- and b-paralogs.

The KCNA2/1 analysis shows also a clear division between these two set of genes, but there are no obvious rate differences between them, only within the teleosts somewhat increased rates are apparent [see Additional file 2]. For both genes, the non-teleost fish sequences (Acipenser, Amia, Lepisosteus) are pro-orthologous of the FSGD as had been proposed in previous studies 103738. Studies based on Hox genes as well as other nuclear genes (sox11, tyrosinase, fzd8, POMC) found a phylogenetic timing of the FSGD after the divergence of Polypteriformes (bichir), Acipenseriformes (sturgeons), Lepistosteidae (gar) and Amiidae (bowfin), but before the teleost radiation including the most basal group of the Osteoglossiformes 1030. The Gnathonemus KCNA2 sequence is positioned basal to the duplication in the gene tree. The Gnathonemus KCNA1 sequence is grouped with the b-paralog in the KCNA1 gene tree. But also in this analysis, statistical support for most of the nodes is lacking.

The KCNA5/10 tree shows a similar pattern as the KCNA3/6 analysis with a clear acceleration of evolutionary rates in the KCNA10 genes, a trend that is even more pronounced in ray-finned fish [see Additional file 3]. The Hydrolagus KCNA5 sequence is phylogenetically grouped with the other KCNA5 genes with good phylogenetic support (94%BP, 100%PP), indicating that the duplication of the clusters leading to KCNA5 and KCNA10 occurred before the divergence of cartilaginous fish as previously proposed 39.

Even though the phylogenetic analyses of KCNA genes cannot pinpoint the duplication event in the fish phylogeny with a high degree of certainty, the numbers of identified genes implies that the phylogenetic split of basal lineages that include Acipenser, Lepisosteus and Amia from the fish stem lineage precedes the duplication event, while duplicated KCNA4 and KCNA10 genes suggest that Osteoglossomorphs (Gnathonemus petersi) diverged after the 3R event. This interpretation is in agreement with previous analyses on the phylogenetic timing of the FSGD 10.

We analyzed the non-coding regions of the complete clusters using the Tracker software 40, which detects clusters of such phylogenetic footprints (putative transcription factor binding sites) termed cliques (FC). Following the definition of phylogenetic footprints as in Tagle et al. 41, we only compared sequences with an additive evolutionary time of at least 250 million years, i.e. species that diverged at least 125 million years ago, and therefore we excluded comparisons of orthologous pufferfish clusters 4243. For the KCNA 6-1-5 comparisons, we included the partial Tetraodon KCNA 5-1b cluster, since it was currently the only available genomic sequence for this paralogon. Because of a large sequence gap in the intergenic region of the medaka KCNA 6-1b, we omitted this sequence from the analysis. We also could not include the Danio KCNA 3-2-10a cluster, since no linkage information for KCNA2a was available in the current assembly. For these analyses we were able to include a total of 20 clusters (11KCNA 3-2-10, 9 KCNA 6-1-5 and we obtained 670 FC of which 182 are shared by more than two species. The alignments and positions of those are given in the Additional file 5. Since an analysis using untreated sequences resulted an unusual high number of cliques shared only between the two paralogous human clusters as well as between the two Xenopus clusters, an effect probably due to the abundance of repetitive elements, we used RepeatMasker to remove those elements and repeated the analyses 44. This strategy reduced the number of intra-species hits successfully.

We also analyzed the number of FCs and the length of conserved sequences between orthologous and paralogous clusters. Between orthologous clusters, the number of conserved elements follows the expected patterns, at least within tetrapods and the orthologous fish clusters respectively (Tables 1, 2). Paralogous fish KCNA-gene clusters share surprisingly few conserved elements.

Interesting in this regard is the comparison between paralogous clusters (e.g. Homo 3-2-10 vs. Homo 6-1-5). While one might expect to find a reduced number of conserved sequences compared to the orthologous comparisons (e.g. Homo 3-2-10 vs. Xenopus 3-2-10), this is not the case for comparisons among tetrapods. The human KCNA3-2-10 cluster shares more FCs with its paralog KCNA6-1-5 than with the more closely related frog KCNA3-2-10 cluster, even after the elimination of repetitive sequences (Table 3). This effect is mainly due to relatively short FCs that are found only in the human sequences and not in any other species.

Up to now, shaker-related voltage-gated potassium channels have been mainly studied in tetrapods with a strong emphasis on functional and structural aspects 343545, but not within a larger phylogenomic framework. Neither the number of genes within ray-finned fish nor the phylogenetic relationships of these genes have been studied previously. Yet, this information clearly provides useful insights for the comparison of experimental and functional studies. With prior knowledge of the existence of two 3-gene-clusters (3-2-10 and 6-1-5) in mammals 3233, we were able identify two clusters in chicken and the frog Xenopus tropicalis as well. The human clusters are positioned on chromosomes (chr 1, chr 12) that have been reported to contain a number of genes duplicated during large scale duplication events 7. The addition of new data from fish reveals the existence of four clusters in teleosts as a result of the fish-specific genome duplication (FSGD). KCNA3-2-10 cluster of Tetraodon are on chromosomes 11 and 9, in Oryzias on chromosomes 7 and 5. The origin of those chromosomes through the FSGD has been proposed previously 1646. Because of missing linkage data, similar conclusions for the KCNA6-1-5 clusters cannot be drawn. Due to a lack of data from lampreys and hagfish, the timing of the first cluster duplication (leading to the two-cluster situation in tetrapods) is unknown (Figure 1). Most likely their origin is the result of one of the genome duplication events during chordate/vertebrate evolution (2R) 747. The two genes we identified from Hydrolagus colliei (spotted ratfish, Chondrichthyes) so far can unambiguously be assigned to their tetrapod orthologs (KCNA2, KCNA5). This finding implies that sharks already possesses the two KCNA clusters 39, the timing of the duplication with respect to the cyclostomes remains unclear. The KCNA complement of the cephalochordate Branchiostoma floridae shows an independent series of gene duplication, with all genes being intronless. Thus cluster formation in the vertebrate lineage occurred after the divergence with the cephalochordates. In Ciona intestinalis, a tunicate, only one KCNA sequence was found, which consisted of at least five exons, of which only four were identified unambiguously. The amino-terminus was found to be highly variable among different vertebrate genes and thus BLAST searches with invertebrate sequences yielded no hits for the amino terminus.

The phylogenetic analysis suggested KCNA7 as the most basal vertebrate KCNA gene. This gene has two exons, while the rest of the vertebrate genes are intronless (Figure 1). The phylogenetic analyses could not resolve all sistergroup relationships between the intronless KCNA genes (KCNA1-6, 10) with high confidence. This might be due to the extreme rate difference among the various members of this gene family as well as, the rapid succession of duplication events. We propose an evolutionary scenario of two consecutive tandem duplications that formed a first cluster. Mostly likely during the 2R genome duplication, the entire cluster was duplicated leading to the situation found in tetrapods (Figure 3). This hypothesis is also supported by the sistergroup relationship found between KCNA5 and 10, as well as between KCNA1 and 2 genes, which are now parts of different KCNA clusters. For the other gene pairs, the data are not as clear, but the alternative scenario of independent tandem duplications on different chromosomes is not supported by the topology of the tree. We suppose that the reconstruction problems are caused by the fast evolution of KCNA6 as well as by the short basal branches that render phylogenetic reconstructions difficult. This also implies that the tandem duplications must have happened, evolutionarily speaking, only shortly before the entire cluster duplications. The origin of the KCNA4 gene from the KCNA7 gene progenitor could be the result of a large-scale duplication (1R) event followed by the loss of several genes but currently there are no synteny data available to add support this hypothesis.

The phylogenetic analyses, also with smaller datasets [see Additional files 1, 2, 3], are hampered by pronounced rate differences between paralogs and increased rates of evolution of KCNA paralogs in the teleost fish. Accelerated rates of evolution in some fish genes have been observed before 484950. Pronounced rate differences can lead to wrong topologies 51. An accelerated rate of evolution might be due to reduced selective pressure because of gene redundancy after duplication, or, alternatively, might be due to positive Darwinian selection associated with a change in function. Yet, currently, a link between expression and the formation of heterodimers between different subtypes and their evolutionary rates is not obvious. More functional data with respect to gene expression and its regulation, as well as the formation of heteromers from more species, especially teleost fishes, could provide further insights. To date, expression data are available only for few tetrapod species such as mouse, human and chicken. Furthermore, these studies only examined expression differences in a subset of tissue types 52.

For a better understanding of the characteristics of the gene clusters, we first performed an alignment based VISTA plot analysis 53. However, no clear conserved regions across all species included were apparent, only between the orthologous fish clusters (results not shown). For a more detailed analysis, we ran the Tracker program, which is based on an initial BlastZ algorithm 40. We performed a complete analysis using 20 sequences and obtained 670 FC, of which 182 consist of more than two sequences. Alignments and sequence positions of the FCs are provided in the supplementary data. The percentage of FCs of the complete sequences was very low in tetrapods but the teleost clusters showed comparable values (11–34%) to the Hox clusters (10–38%, Hoegg et al, submitted) (Table 4). The lack of signal observed in Vista plots is rather a pronounced lack of elements conserved between tetrapods and fish. Hox genes as developmental key factors, on the other hand, are involved in many different pathways and, therefore, have more regulatory elements that are conserved over long evolutionary distances. Expression of KCNA genes, on the other hand, was studied so far only in adult tissues, but a more continuous expression pattern in specific tissues (e.g. brain, heart, muscle) might be expected. Since KCNA genes are active as homo- and heterotetramers, expressional "fine tuning" might also be accomplished through the expression of a combination of different KCNA genes at different stages during development and in different tissue types. The analyses of orthologous clusters revealed the expected pattern for the tetrapods and within orthologous fish clusters, i.e., more closely related organisms share more cliques, while the paralogous fish clusters share less conserved elements (Tables 1, 2). This finding again is most likely caused by faster evolution in duplicated fish clusters.

However, the comparisons between paralogous clusters of the same species showed an unexpected pattern for the tetrapods included in this study (Table 3). The phylogenetic analyses show that the clusters were duplicated before the tetrapod – ray-finned fish split (+/- 450 million years ago 54). Using sequences that were not "masked" for repetitive sequences in the Tracker analyses as proposed in the original publication 40, leads to high numbers of FCs shared only between paralogous sequences of one tetrapod species. The teleost KCNA clusters are more compact, contain less repetitive sequences and are therefore less susceptible to false hits. For gnathostome Hox clusters, where selection is acting against repetitive sequences 55, masking can be neglected, but for other gene clusters it can be useful.

This finding might imply that the clustered structure of KCNA genes is to some extent due to common expression domains of the genes within each cluster 56, or at least that the regulatory elements are not conserved over larger evolutionary distances. Since expression data is not widely available especially not for non-mammal species, it is difficult to draw further conclusions about this testable hypothesis.

The KCNA gene family underwent a series of duplications within the vertebrates leading to eight genes in tetrapods and more than 13 in teleosts. The initial gene underwent at least two tandem duplications forming two three-gene-clusters, that were most likely duplicated during a genome duplication before the divergence of the Chondrichthyes. The molecular evolutionary analysis of the KCNA gene clusters showed only few footprint cliques (FC) that are conserved over large evolutionary distances, while among the teleost clusters more conservation was evident. Shared regulatory elements do not seem to be the major force that keep these clusters intact and therefore do not pose a general rule for gene cluster retention.

SH participated in the design of the study, carried out lab work and drafted the manuscript. AM participated in the design of the study and the writing of the manuscript.